how random do you think you are a few months ago I made this rock paper scissors engine to answer that question and how it works is simple it starts off by having absolutely no clue what the hell you're going to play but as you play more and more it begins building a psychological profile off of your moves to better predict what you playing return and before you know it after a 100 or so games against it if I just move this out of the way it begins uncovering these unconscious patterns you have to suddenly go with certain choices and it rates these Trends with confidence values so then after a while of playing against it actually becomes almost scarily good at predicting your moves so this got me thinking can I somehow recreate this algorithm in a far more advanced way to go from playing rock paper scissors to reading your mind or at least trying to here think of this if I ask you right now to think of a number ranging from 0 to 100 your number may be I don't know 13 94 65 or literally anything inside this range so normally you would think oh for me to guess what number you're thinking of right now the odds of that are super damn low and you would be right until I found this [ __ ] study about human Randomness and it turns out when humans try being random they end up being predictable simple-minded monkeys which I found to be pretty interesting but now I wonder this study only encapsulated the first number that came into their heads What If instead they persisted coming up with numbers just how much psychological gauging and algorithmic training would it take to then create a learning model that is able to predict the human mind and with that even able to guess accurately more than 1% of the time I don't know but in this video I'm going to be trying to get that as high as possible Jesus Christ this video is so damn stupid anyways before I begin trying to build Skynet in the rise of the Terminators I quickly want to switch back to my basic rock paper scissors code to explain what the hell is going on when the game begins you are presented with three buttons rock paper or scissors now upon originally loading it up the engine has no information about who you are and how you play so just for the first move I made it always play paper because with humans Rock has already been recorded to be the most popular First Choice so statistically it's already at a slight Advantage but what it's now done is record that move that you just played in a list and every move that you play from then on also gets added to this list but after a good enough sample size is recorded something interesting happens it begins trying to identify patterns by going through each index of the list and asking itself hm is this spot here from the past the same thing as what's going on right now in the present huh oh crap it is the same but how similar is it though well the pattern has been exactly identical for the past five moves straight so surely the next move that's going to be played is scissors right no crap even a first grader could find that pattern but what's interesting with this algorithm is it doesn't just stop and play the counter to scissors right here and now no no no no no that's too easy to ensure its accuracy it continues to go through each and every index of the list questioning what the hell the most prominent trend is just like over here more recently we have the exact same pattern as before but what's different is in this Clump that was played next wasn't scissors I broke the trend and played paper instead so how the hell do we decide whether we should play the counter to scissors or paper and the solution is appeal to human psychology humans struggle with retroactive interference which means the more recent thought patterns tend to be more replicable than the older thought patterns think of retroactive interference like how right now if you were to go on your phone and type in your password it would come into your mind instinctively because of just how often you've been doing it recently but then if I asked you to type in the password that you had 4 years ago you go into a coma and have no damn clue what your password was back then this is the fundamental aspect of retroactive interference but the same thing could also apply when you're rapidly playing rock paper scissors as well so if this pattern here was super old and this pattern over here was super new we want to Value this newer data more and then we just have to store that for how confident we are for each value but these two aren't the only patterns that we could find there's also a bunch of random smaller little patterns that might influence the decision and after adding all those up these are the final confidences the bot things I'll play so then all we have to do is take the one with the highest confidence and then play the counter for that move and it does this process after every single move you play against it over time just getting stronger and stronger at predicting you so now that you understand this base concept I want to see just how stupid I could get this algorithm to be but to start I made this little window here which is going to be the main way to enter numbers currently this window is very Bland I know shut up I'll try making it better later but for now the app currently only has two functions and that is by using this text box here we can enter a number and what it's looking for right now is only numbers between 0o and 100 so if we enter a number inside this range it will display a text that shows a predicted number and because it currently has no brain and it doesn't know what to predict I just made it say kill yourself constantly so there's that and if we enter a number that's outside this range it won't even accept the input and what's neat is you could enter numbers pretty quickly by just hitting enter and not using that stupid button anyways enough of this Bland SGI it's time to move on to the [Music] [Music] code or in other words I just spent an entire few hours binge watching Breaking Bad while typing random things into the code here and there and it was very productive basically what it does now similar to how the rock paper scissor spot works I first started by just comparing random spots from earlier in the inputed list to the more recent inputs that I've done to guess what will be next only difference being is when I was working on the rock paper scissor spot I manually hardcoded its ability to look five moves back and compare it each time and that system is garbage humans though being stupid as hell still do have patterns but sometimes they are far far less reoccurring than in groups of just five so to combat this I made this goofy system here which means I can now have as much depth as I want and uh I think a 1 million should do the trick so our pattern prediction can now be so vag we can grab the entire be movie script of 13,760 words convert the entire thing into Morse code which puts us up to 176,000 characters and then convert that for a zero for each Dot and one for each line and then make that so once every five playthroughs of this entire binary be movie script we just add one more one at the end of all of it it would still be able to remember all those ones and zeros after just receiving 881 th000 bits of information H does this work well it works by looking at the confidences for all numberers 0 through 100 and based on the sum for how strong each trend is it adds how confident it is in playing that number so if we have the entire B movie script over here with five playthroughs for it to learn and with that random one added to the end of it and after seeing that that's there once every five Loops it's going to feel Mighty [ __ ] confident that that Trend will continue and also for those weirdos who decide to inter numbers and arithmetic and geometric patterns I got this bunch of code here that says screw you and predicts of that because it is very easy to do but another thing I want to add is as you can see right now I'm entering a numbers super quickly it's kind of hard to see if the bot actually got it right or wrong every single time so I quickly want to add a little something that plays a sound if it gets it right this folder here is all the sound effects I use for my videos let's see if I could find something here ah how about this Jesus Christ that's loud oh hell no Ah that's pretty good but I don't think it's affirming enough what about a ding nah oh wait this should be good all right all right and it has a failure noise wait that's perfect I could definitely use that even though this is the same sound effect that plays in those cringey kids YouTube family channels have an Apple Watch and a regular one all right got to trim the edges all right there's the correct sound effect and yep there's the incorrect sound save those bad boys and then get it to play real quick oh this is about to be good okay first number one okay no two no one two one okay okay got it and then two again all right there we go it actually gets it oh my God it's actually working this is going to become so annoying so damn fast I love it okay I believe it's time for the first test to see how good the algorithm is so far so here I just got done adding three metrics which are going to help try measuring that first is the total win rate 100% meaning that the bot has correctly been able to predict every single number I've entered while 0% means I hasn't got a single one correct so we want this number to be high but I also don't want to cheat the other thing I added is the actual confidence table for each value it only updates the confidences after you submitted the prior number or else that would be cheating but it's there to see if it's on the right track and the last thing is the amount of rounds played so in order to give this a good test I want to start putting in numbers that are actually good random numbers and a lot of them this process might take a bit so I began putting in a bunch of numbers and to be honest I didn't have super super high hopes with this model so far but as I kept going on while still maintaining a good level of Randomness the bot was actually able to get a few correct which caught me off guard but the main issue I began noticing was the more I played against it was it began taking super damn long to generate each guess this is because as I was recording this I forgot to break out the damn Loop so the code is very right by input 500 I already got so bored I took off my headset and began listening to music on my phone I also realized a bit later to decrease the time in between inputs I began backlogging some the numbers in the text box while I waited but holy Jesus Christ did it still take a while I even got so bored I began trying to stream movies on my second monor when this happened oh oh [ __ ] oh no no no no pause that real quick don't you dare crash no don't oh my God thank God oh my oh god I've never been so happy to hear a buzzer in my life okay there I was able to put a 97 numbers into the bot and look I'm sorry I couldn't get to a clean 1,000 you try spending two goddamn hours typing nothing but numbers anyways with all that complete we got a total accuracy of 6.6% no I'm just kidding uh 6.6% isn't that good but it's higher than 1% which means mathematically there is some predictability in my inputs so now it's time to try and exploit that to the max after doing that test my main conclusion for to make it better is to probably stop treating each number as an individual Instinct here's what I'm thinking currently the AI is a a master at finding strict patterns and strict patterns only but humans aren't always like that in this pattern above once the bot is trained on this pattern it will have no problem repeating it again and again but the issue is if we swap out just one number here to a three instead of a two this is what the AI thinks is going on and that's not good and now totally discards this recognized pattern from before just because it doesn't perfectly align with this new one so to get around around this I began thinking most people who are inputting numbers probably have a keyboard like this sorted by counting up with the exception of the zero being at the end and I believe this may affect the way people both conceptualize random numbers and type random numbers so what if instead of just analyzing the pattern and this pattern only we began trying to treat the keyboard as a normal distribution graph and we begin rewarding the keys closer to the actual study guess for both the first digit and the second digit so subconsciously patterns can become more generalized this also means we can begin finding more subtle patterns to user inputs such as taking apart each digit for each entry in the inputed list and put that into new lists that we could study then by learning these patterns we can then better predict whether the user's chubby hands tend to slide up or down the keyboard for each integer and or do they just tend to stay relatively close if they do end up moving even if it's subtle all we have to do is guess the most likely location their hands will end up on on the keyboard then we use the normal distribution to reward the first digit confidence in each number then a bunch more distributions to reward the second digit confidence within each 10 number bucket this here is cool and all it's an interesting idea but it still has one issue and that is if I don't want the AI to be completely stupid every single time the application is launched I would need to train the algorithm beforehand so then the algorithm is already pretty smart at predicting people before the main user even begins inputting their own numbers only problem is if I want to train the AI accurately I'm going to need a lot of data a a lot of how do I get data oh wait a second [Music] so it's been about 5 days since I launched my Discord bot and began stealing data from you guys consenti of course and holy [ __ ] did I get a lot of data you guys are really that damn bored that you guys started forming movements and Cults trying to get this number over 10,000 now it's at 21,000 but oh my God there's even a point where the bot shut off and died for like 5 hours and you guys held a goddamn funeral for it what is wrong with you guys anyways you may be wondering why I'm doing an IRL cam right now and that's because in order to keep my bot online I had to have my PC turned on 24/7 and was I really about to do that when I'm sleeping right here no no I don't think so that would keep me up more than my thoughts do already so instead over here I have an alternate computer and this computer has been a powering the bot ever since I started it this VC back here is very good it's super quiet has good specs and is super damn hot cuz it has no fans in it you could burn yourself just by touching it but it's whatever it worked anyways how I've been storing all this is after any message gets sent if the message is between 0 through 100 digit then it gets appended to this giant ass file this huge txt file with all your data and what's neat is after any number gets sent it keeps track of who sent it so I could keep their individual data together this could be useful for psychological profiling so uh thanks of the vergins on my Discord I have your data now so what the hell does one do with over 21,000 now sorted handtyped numbers well we Cherry Picked the hell out of it because half you damn idiots just decid to count which I specifically said in the instructions is doooo because I already have an algorithm for that anyways what we can do now for every number in the data set is we could look at every following number that was typed after then we could gather this for each number and then split it for first and second digit from there if we sort it from lowest to highest we are then able to find the median following integer for each number then we can file all that into a dictionary so it's easily callable later and remember what I said earlier about just predicting the user's input for each integer well we could try doing this by using a random Force regression model which is this [ __ ] thing which could be used to predict sequences and then all we need to do is run that twice to predict both the next first digit and second digit and from what I see it's actually pretty dang good at predicting numbers so to combine these two ideas to find a more accurate Middle Ground round all we have to do is gruesomely split it in half to find the average then I worked on this pretty neat normal distribution algorithm going on here which basically accepts a Target first digit and second digit that it want to reward so let's say the number we want to reward is 24 when it gets ran you can see it spits out a whole lot of rewards but mostly gives it the points of 24 but it also still rewards the neighboring fours like this one down here and all the way up here and also rewards the numbers that share a closer first digit so the further away the numbers are from the target digits the less so get rewarded holy crap that was a lot of explaining but I just got done putting all that in and I'm only fairly sure I didn't drastically screw everything up so it's time to test it so here it is but you may be thinking oh how are I actually going to give it a proper test to know if it actually improved or not because of blah blah blah blah blah shut up I thought ahead before I did the last test I put in a piece of code that made it print the full inputed list into the terminal after every input so all I had to do was take that same list of numbers later and program this button here which I conveniently decide to call the 9 07 test button to call out how much of a lazy ass I am for not hitting a th000 when I did it anyways how the 907 button works is by pressing it it begins repeatedly going through and quickly putting in the exact numbers from before holy [ __ ] this is loud what I'm trying to say is by giving each version of the ey the exact same standardized test I can see if it actually improved also I've greatly improved the speed of the AI so it should actually be finishing quite soon oh oh wait it went down why the [ __ ] did it go down I I I did so much hold on hold on surely I I got to change some values what if I try it again without a data set what if I try it with a different amount of in estimators what if I change the standard distribution what if I made it rely more on the data set why won't this work what the [ __ ] are all these regressors it's impossible it's impossible to get this number to 10 I got to keep track of any minor improvements keep going up keep going up oh I got to make more snapshots of the trade models yay how many more Little Numbers do I have to change huh that did Jack [ __ ] there's no way to boost this win rate there's literally no way I can't reach 10 it's impossible to reach 10 it can't be done it literally can't be done what the [ __ ] am I doing with my life [Music] it has now been achieved a win rate of 11.24 6% and yes those decimal points are very important for many of you that may not seem like a lot but it really is this means out of the 97 samples my bot was given for this test it is able to exactly correctly guess over 100 of them consistently that means if at one point there was an input of 32 the AI would have to also perfectly guess the exact 32 for it to be counted as a win meaning it could just be one off and it would be counted as a loss instead so realistically there's a one and 10 101 chance that it gets it correct which is equivalent to a little less than 1% due to 0er and 100 also being eligible numbers and my bot still never being exposed to these numbers anywhere before the test was still able to score over 1,100% better than that and even when it doesn't get it exactly spot on it still does tend to be very close to the actual number but working on this has been no easy task for nearly 2 weeks straight I've done absolutely nothing besides waking up and meticulously scraping through my code racking up dozens of hours of footage I even at one point got so desperate I set up an automator to begin running thousands of 9907 tests in my sleep with each One automatically making a small micro adjustment to one of the weights just to boost the win rate by a fraction I even tried lending this project to a math wizard to see if we can increase the win rate the results ended up leaving her insane and she didn't even raise it by Z .1% there hasn't been a single minute in the past 2 weeks where this audio wasn't constantly playing in my head and do you want to know the main secret on how I was able to nearly double the starting win rate well the secret was brutal hard work and self-development yeah I'm full of [ __ ] you see all these people here they made a [ __ ] ton of regressor algorithms throughout history so what the AI does now is by feeding the necessary data to these learning models they can return back and estimate for the the first digit and second digit but also a full number as well now alone these estimates tend to be garbage but by using my normal distribution idea I'm able to layer together the distributions for all the first and second digits and then on top that layer in a predicted full number guess as well with less standard deviation and remember what I said earlier about the median following number dictionary and how we're going to need to average that with the regressors yeah terrible [ __ ] idea instead we could also just pour that in separately so even if each estimated number by itself is terrible by just making a lot of garbage guesses they're eventually able to find a consensus then unless there's a subtle strict pattern or mathematical sequence recognized we just return the top number so what my job has been is to carefully tweak the hell out of these regressors and find out which ones tend to be more ass so I'm able to make them have less of a say on the guest number and by screwing around with these algorithms so much I've been beginning to learn a lot about how humans tend to conceptualize Randomness and the results I found were when Under Pressure human creativity turns to [ __ ] and I could prove this by looking more in depth into my 9007 test now if you are listening from before you know the earlier versions of my number guesser were extremely unoptimized so unintentionally because the time in between number insertions became steadily slower as time went on my numbers actually also became more random what you're seeing right here is a line graph where each point represents a spot in the test in these spots I fed only the neighboring 50 numbers to the AI so 101 numbers total and then I let it clear its memory every single time before gauging the next location so it was around here in the test when when my method to input numbers was to type one on my PC lay on my bed for like 40 seconds watching Hannibal quickly come back and type another number and then continue watching my show and I believe it's because of that buffer space in between typing numbers which led to the predictability of my numbers waterfalling I can only assume this is because my subconscious state of mind was more refreshed after watching the show and not only that but my hands themselves were also likely placed in a more unique position each time I return so it didn't have that prior number I just entered influencing where I may want to go but in this beginning section I was absolutely destroying my keyboard entering numbers quickly and you could also see the Bots relative win rate around this spot was also higher but now you must be stinking to yourself right now oh um I I wouldn't do something like that I'm not predictable I'm built different no the [ __ ] you aren't well maybe you are my 907 test did actually have some giant artificial spikes occasionally because I did sprinkle in a few sequential numbers to see if it'll catch it sorry about that but I really do think that this AI can be used as a metric to quantify creat it and predictability among people so by keeping in mind that whole pressure messes with creativity idea I added some code here which now when the program is launched there is a clock and this clock is essentially just a time bomb so right after you send your first number it begins counting down and as you can see it works perfectly in the background also I can easily edit the time when it starts and finishes here so when I run it at 5 seconds and wait for that I can no longer send numbers once it's expired so now in entering numbers people are constantly under that good old press pressure I need them to be to be nice and predictable and what exactly am I planning to do with all this well now I'm wondering is my crappy artificial intelligence spot somehow smarter than actual human intelligence and uh now you may be wondering how the [ __ ] are you going to be able to test that simple I hit up my good friend Beast spot mentioning a time when I want him to be available to record and also some money to make him blinded for possible DB a then I spend an hour developing this new app which as you can see has this little text box and the objective of this app is to guess what numbers are in the sequence see I just got 15 right because I know it's the first number in this list but this will be hidden later so if I type something shitty it marks it wrong because the actual answer is 20 so if I spam it real quick it allows you to keep track of the past 10 numbers and whether you got them right or wrong and if you keep guessing numbers eventually you reach the end and it tells you what your final win rate is then we add to my other friend Tyler telling him that we're going to meet on the same day as bpot then all we have to do is convert the main app into an exe file using pie installer which should be a pretty simple process uh what do you mean main isn't found you're literally in main no no oh my God this is so annoying please Indian tutorial save me all right oh shoot that actually worked all right it wants to file so I'm going grab them put them here all right then launch it it isn't even loading what the [ __ ] was that that did nothing wait no n screw turning this into an executable file I'm just going to [Music] uh Dom things you're either gay for me or you have cancer what is it so I'm going to send my entire bot over to you and I want you to play it okay you should screen share by the way cool here let me just send you the file I don't know if like that would be faster just me doing this there you go now just download on that yeah just click that are you white screening from downloading that oh uh no there is this all the numbers you got uh yeah that's the entire data set it's all in one file okay yeah I'm going to send you some stuff you're going to need to like put it in like stick it in yeah yeah the path it's you'll see like so download this and [Music] this no no stop I hear that in my dreams oh okay it finally worked all right there and now move it to the right so what the objective is I want you to get as low as a win rate as possible Right AK trying to be as unpredictable as possible and you have to type in 500 numbers in the 11 minutes of course Jesus Christ that's loud yeah the time started it should start taking down it does require a lot of uh keyboard spamming are you just typ bro yeah you just start typing what it was putting it's obviously going to catch that you can't fool it that easily I don't know what I was doing I'm just putting in numbers oh got him there what is that laugh oh I got it right again oh 23 in a row oh one more time free yeah the win rate is going up for the bot it's learning you is it bad that I'm locked in uh yeah you suck 23 and then if it [ __ ] on me I do it did 32 when you typed 20 okay it got the number swapped oh now what2 watch it 51 it's not going to free I was just going to like I was Ed I was edging it you were edging the AI uh oh oh I almost did a yo it almost got me there you're trying to outsmart they are mhm bro that's so random freak just chose random number OT yeah you're right I I I'm using random numbers you see if it was truly random like it would only get like a 1% win rate a little bit less than that but it's not that's H him with that 23 it's all that bro you're not you're not fooling it dude okay maybe you just learn yourself oh it was a one off there it was one off all right now I got it 23 g you don't have to dude chill out it's working mind is it you only have like a minute 25 left I know that's why I'm doing this cuz this is a Qui come on all right but don't don't kill your keyboard I don't care about my keyboard yo I hear your like little fingers prancing around 10 left 10 left all right this 10 you can actually probably take your time light work light work you know what I got one more what's your last number going to be I did it on 23 it got a 5.2% win rate against which is actually very good at the end I was spamming so all right so you see in terminal there you see all those numbers just inputed yeah these ones copy paste all those to me down a little bit yeah you should see it like ends with a bracket just give me all those numbers there okay I think my predictability went like up a lot when I started spamming because like I like it's not like I was like moving across my entire keyboard I just kind of yep and that was the entire point I was also just trying to add social pressure that the entire point of it was to actually pressure you anyway is Tyler was actually like never actually like gone and I'm actually going to add him to a DM real quick there create group DM all right he's here yo before i y turn on your camera right now Tyler I need my camera this is the safety Proto um hold on let me plug my phone in that oh he's sexy and he's beautiful guys all right what is what is happening dude so I just got B Spot to like enter like a billion numbers AK 500 and I'm basically just going to screen share you're going to have to try guessing it from my computer basically and if I see you like I know you have two monitors if I see you turn like away and like bot is texting you and like giving you the numbers that he just sent me I'll know you're cheating okay okay all right I'm going hide this his numbers there I got it all loaded up here is the app so every number that he puts it's going to be on this like giant column thing if you're looking so you can like see like what what's he going to do next so bpot got a 5.2% win right sure and if you could get higher than that I'll give you both $20 so $40 total you mean lower than that or no he has to guess it better than the bot basically okay I see yeah yeah there's 500 remaining the final win rate gets shown once you're only done guessing all 500 what number do you think bbot put in first and I'll just type it what number what number do I think you put in first [ __ ] think you be can I contemplate yeah yeah dude take your time I'll start with uh zero he gets 23 for his first number and you can see it got it wrong there will be a red and green if he gets it right or wrong I like it you know I I have a feeling there's a logical pattern here um 23 you went to 65 [ __ ] this kid I'm just going to start like Fast balling it just like yeah yeah you're going to need to do a bit of that there's like 492 left can you do the numbers 0o through 50 consecutively what do you mean like 0o 1 2 3 4 5 6 7 you want me to do that right now you got one [Music] there yo so you wanted 100 as well yeah okay Jesus bro all right are you going to keep automating guesses or are you just going to like uh let's do uh a countdown from 100 to zero you're not for real oh I'm I'm for real yeah you you know you can put more thought into it it's random well it's bpot random it's guessable this is bpot okay whatever based off of your reaction there's meant to be some discernable pattern that I'm meant to be analyzing here and you were hoping some analytical that I would like look and find something it's very difficult it's very difficult I mean honestly a good strategy would be to just pick up on one that he types A lot and then just do that like a lot and then that one right yeah I know he likes there's two in the' 70s so we we could try doing just a bunch of things in the 70s so if you want to just do for the next few turns go from 71 to 79 uh no I don't he doesn't usually do things with the zero in it so I would do 71 oh immediately 72 all you know what just keep doing 17 for a while I'm not even joking [ __ ] [ __ ] 71 that's the same thing just keep doing 17 and I I don't even care oh my God this is aggravating uh you want me to stop you feel like you're wasting these guesses I no no no do another do another 17 okay I feel like a gambler yeah the the the pattern here is that he does seven a lot I I know I know I mean I could try to be more analytical but considering he was being random I'd probably be overanalyzing course you're overfitting your own mind yeah exactly I don't want to overfit here the Beast bot nurdle Network it only has like three nodes Max yeah it's just input to Output nodes there's nothing in between no hidden layers we're such [ __ ] nerds nobody watching going to know what the [ __ ] any of that means I don't know what you're talking about so what I'm noticing I guess the the main pattern is he's not hitting any of the numbers in the middle and his trying to be random I assume while he was doing this he was doing a really quick succession at this point so he's doing all the numbers at this end and then going all the way back here he's like going here then here then here then here going on a 90 streak oh yeah a 60 that makes sense uh go for nine let's go five five seven four8 n actually don't make me lose $40 like I know it's funny to begin with but don't actually oh yeah yeah out of your Millions dud um $20 is a lot bro $20 $20 I like how you'd have to blur all the footage if for the rest of the video I was like um oh that's crazy huh how does it taste I don't just blured I'm like man I don't know what to put what should I put guys what should I put go go do uh 10 I hate you spot do a three then five yeah there's no [ __ ] he's just it's just the monkey with the [ __ ] like symbols in his brain sh all right I'm going just do like my own algorithm here so you did one now do nine7 then two seven then eight okay now do three you see it's funny cuz like you're thinking logically in the same way I would think but like yeah if you couldn't tell what I was doing was I was just working my way towards the center from both so one and then I well but when I'm saying second my my mom's texted are you looking bro be stop cheating stop looking at your phone I texted his number give him all the next numbers my mom is giving me the answers right now you're not getting the $20 I've had it enough uh do a six yo this is actually just right do a uh three do um dude what is this footage four you're actually just autistic like what is this for the last 23 guesses I want you to put the number guest that we have left you're not for real actually do I actually want this for your last bit I don't know my win rate so I might sell it let me let me think for a second I'm going to pull out my calculator actually cheating he's using a you have to get 26 correct if I'm doing that Mental Math correctly or 27 maybe whatever somewhere around there no I kind of want 20 bucks I'll go go for four one off all right then I'll go one off from what I think the next one will be go for one so one what I meant was [ __ ] dude [ __ ] well I think it's two so let me go one less than I think it is I think I think now I think now it's uh time to do the uh plan of action go for eight I already knew go for 7 6 5 4 3 2 1 yeah hey yo four [ __ ] three 1.8 damn Bly actually did do a good job at being random there cuz like I was like damn if the bot even the bot got like 5% like that's that's kind of crazy and then like human gets 1.8 like damn good job building a bot dude yeah yeah there goes down two months of my life for no goddamn reason I'm so done with this video so there it is my stupid robot can predict humans nearly three times better than humans could predict themselves which is pretty interesting but now you may be wondering why isn't this video a Minecraft video well I actually want to talk to you about that I've been doing nearly nothing else but Minecraft content for the past four years of my life and with that I've only began doing creational type videos a year and a half ago but I feel like I've reached a roof cool let's do an entire mojing update in one week all right now let's create Discord in Minecraft nice now a 3D printer all right all right now create a parkour pathing algorithm oh okay now let's create a literal chess engine inside of Minecraft what the hell but now I wonder how do you top that and still be able to post within the next millennium this is why I feel like I've reached a realistic top for how far I could go with Minecraft content so that's why I finally decided to step away and make this video this is currently the first video of its kind so I bet there's a bunch of ways I can improve and uh not make it garbage but with this change I hope I could finally start putting out videos more often and not just any videos but videos I'd want to watch myself because I don't only want to be able to post one video every 3 months that's terrible and you know what I'm terrible for doing such a thing so I'm sorry and as an apology I would like to make it one month instead is that is that how you end it I have a better ending for your video