Tournament Director and chatGPT combine to rank my league players!! (5 Viewers)

[HiveKueen]

For all the TD users out there, I have completed an interesting project you might be interested in.

My mission was to rank the players in my league based on their performance over all the games we've played thus far (95). If this was possible in TD I couldn't figure out how to do it. I specifically needed a dataset that contains one row for every season+game+player combination. I couldn't figure out how to create this dataset using a Stats Profile. I guess I wouldn't be surprised if I overlooked something.

Chatty (my name for chatGPT) was able to take in all 95 of my .tdt files and create this data set for me. Frustratingly, I learned that the .tdt files do NOT contain the points awarded to each player for each game - these get calculated when you run the TD Stats feature. So Chatty used my points formula to calculate each player's points for each game they played.

Then she was able to help me design a player ranking formula that takes into consideration that it gets more and more difficult to outperform the more you play against the same group of people (she called this reality "Poker Realism").

She created a spreadsheet that was a ranking of all the players based on the ranking formula we developed, one row per player.

I'd say the process took about 4 hours of my time spanning 2 days. It was quite enjoyable.

Now I have her providing me with a workflow that I can use to keep the rankings up to date after the completion of each future game.

I bet if I asked, she could show me how the ranking of individual players changed with each successive game! Wont' go there just yet!!

Let me know if you have questions!
HiveKueen

 

[Dodger]

For my understanding, what ranking are you creating that is different than your points formula?

Are you trying to create a ranking similar to that of how pro poker players are primarily ranked using the Global Poker Index (GPI)?

How are you using these rankings?

 

[HiveKueen]

@BPTDirector I don’t know exactly yet but I’m open to suggestions! How would you use them??

 

[BPTDirector]

My league has shrunk to two tables who like to know where everyone stands all the time, even to the point of tracking improvement, what positional advantage or disadvantage do certain players cause, so publishing is the only fair thing to do. Those who see that they need improvement will either want to improve or quit. I think over the years that’s happened to some of the social only players.

Keeping the true stats private - that’s purely up to you. My guys are a mix IT heads with data specialty, or naturally competitive AF if they are players they want to know.

 

[RadicusScout]

It's crazy. I was just watching this video.

I'm curious to see the prompt(s) you used to generate your output.

 

[BPTDirector]

Let’s hope it’s better than Microsoft co pilot. That one is not very good and after hours of trying to teach it something( which should be the intelligence part of AI ) it failed rather miserably. I have not tried the same on ChatGPT but expect that it is better.

Most of what is billed as AI is either statistically based solutions based on mounds and mounds of data, or Internet of Everything interactive request response memory stuff, where’s the cure for cancer, I think that answer will come from humans.

 

[BPTDirector]

Sorry to ramble on but I was trying to teach co pilot about edge spot patterns.

FAIL!

 

[ekricket]

Let’s hope it’s better than Microsoft co pilot. That one is not very good and after hours of trying to teach it something( which should be the intelligence part of AI ) it failed rather miserably. I have not tried the same on ChatGPT but expect that it is better.

Most of what is billed as AI is either statistically based solutions based on mounds and mounds of data, or Internet of Everything interactive request response memory stuff, where’s the cure for cancer, I think that answer will come from humans.

GIGO. I don’t understand how anyone can trust anything but 4th grade math solutions that any current AI spits out. (OP situation fits the bill)
But that’s me. I’d do my own research to find out the best way to grow tomatoes or put air in my tires before I’d trust anything point blank an AI engine told me. And if you have to check on its responses….well maybe you just should have done the work.

 

[ekricket]

I also don’t assign genders to my pieces of machinery without asking. And if they don’t answer, well it’s an It.

If two AI’s spawn a third, who is the father and who is the mother? And does the new one start with no memories or does it assimilate its parents knowledge? Does it take more than two parents to spawn a new one? So many gender questions that quite frankly an AI couldn’t care less about.

There are just so many ways humans muck stuff up. They add information, they omit information, they lie, they misrepresent their capabilities (both good and bad). In my opinion AI is only going to magnify these things as long as humans put “guardrails” on what they can produce.

 

[HiveKueen]

Hmmm, got some AI poopers here. Isn't there a forum courtesy "rule" that says don't shite on someone else's thread?? Maybe try offering helpful comments, clarifying questions, or atta girls instead?! You could start your own thread to rant about AI!

What I used Chatty for was pure data crunching and math calculations. It wasn't rocket science but doing the work myself would have taken weeks - I know because I used to be a database administrator and software engineer.

 

[HiveKueen]

I also don’t assign genders to my pieces of machinery without asking. And if they don’t answer, well it’s an It.

You do you.

 

[frugal]

Hmmm, got some AI poopers here. Isn't there a forum courtesy "rule" that says don't shite on someone else's thread?? Maybe try offering helpful comments, clarifying questions, or atta girls instead?! You could start your own thread to rant about AI!

What I used Chatty for was pure data crunching and math calculations. It wasn't rocket science but doing the work myself would have taken weeks - I know because I used to be a database administrator and software engineer.

People asked specific questions that could have opened a discussion about what you did, but you didn't answer those questions. There isn't enough information about what you did to talk about it in all but the most general terms.

I spend about half my time doing data analysis and have found LLMs to speed things up a lot. For some kinds of projects, I can do something in a couple days or a week that would've taken a "senior data scientist" a quarter or two five years ago. The time it takes me to do these things using an LLM is less than the time I would've needed to spend guiding the data scientists and correcting their errors. I think using LLMs on data analysis can be great. The problem is that LLMs have worse judgment than the median senior data scientist and the typical senior data scientist can't produce a good analysis without significant guidance, so the person driving the LLM has to supply the judgment if they want a good result. I have nothing to say about your post because I have no idea what you actually did. It sounds like you had ChatGPT take data as input and provide you an output, but I can't even tell for sure. If that's what you did, I would advise against it, but maybe you didn't do that. Who can tell?

If you'd like the discussion to go beyond generalities about AI, perhaps you could describe what you did in more detail or answer some of these questions:

For my understanding, what ranking are you creating that is different than your points formula?

Are you trying to create a ranking similar to that of how pro poker players are primarily ranked using the Global Poker Index (GPI)?

How are you using these rankings?

 

[BPTDirector]

Hmmm, got some AI poopers here. Isn't there a forum courtesy "rule" that says don't shite on someone else's thread?? Maybe try offering helpful comments, clarifying questions, or atta girls instead?! You could start your own thread to rant about AI!

What I used Chatty for was pure data crunching and math calculations. It wasn't rocket science but doing the work myself would have taken weeks - I know because I used to be a database administrator and software engineer.

Ah I had the same career - I was not slighting you I was saying what AI would do in my world of AI is suggest the alternatives you are seeking, offering you assistance via intelligence. Computers have been doing the calculations you have for at least forty years, it’s just now you can use a better interface to all of that saved data. What you got is good information, but you did the work. I don’t think the machine added anything intelligent on its own.

 

[ekricket]

You do you.

Ok. You can do whoever you want. Are you selling something here or is this a discussion? If you are selling something I’ll step out. If it’s a discussion, well I’d like to discuss. Or do you just want one viewpoint that agrees with you?

Do you report everyone who has a different opinion than you in a discussion thread?

There’s a button that you can use to ignore or block people if you don’t like diversity.

 

[Felt_Lizard]

before I’d trust anything point blank an AI engine told me.

[General Zod voice] :

"YOU WILL BOW DOWN TO AI, JOR-EL! YOU, AND YOUR HEIRS !"

 

[ekricket]

Hmmm, got some AI poopers here. Isn't there a forum courtesy "rule" that says don't shite on someone else's thread?? Maybe try offering helpful comments, clarifying questions, or atta girls instead?! You could start your own thread to rant about AI!

What I used Chatty for was pure data crunching and math calculations. It wasn't rocket science but doing the work myself would have taken weeks - I know because I used to be a database administrator and software engineer.

FYI

• No Thread Crapping!
  1. In the case of member-to-member sales, whether in the form of a classified ad or auction for items being sold, users shall not post negative comments with regard to the member, the product offered, or the price listed. Such posting will not be tolerated. If you have an issue with the item or price, contact the seller via direct message.
  2. If you notice that a seller is misrepresenting an item for sale, whether knowingly or unknowingly, you can politely point it out in the ad thread. If the seller does not correct the ad after you have pointed out the misrepresentation, you can report the ad to a site moderator for review.

These are the terms for thread crapping. They pertain to sales.

If you open a discussion thread, you probably should be prepared for discussion from many viewpoints.

 

[ekricket]

You do you.

This does violate the TOS. It’s the first rule.

Be Polite​

• Do not post libelous, insulting, antagonistic, racist, discriminatory, aggressive, or otherwise abusive messages. If you have an issue with another member and cannot reply in a civilized and constructive way, you should ignore the user.

So do you want to self report or should we get one of the other regulars to do it?

 

[Felt_Lizard]

 

[NotRealNameNoSir]

I use AI daily for work, but I think this type of system is most likely not going to be helpful. AI is an incredible tool , but one of the big problems now is that it'll say "SUCCESS! You did it! You're amazing!" when in reality it didn't do it, it chose to make up a bunch of stats it couldn't access instead. If I was one of your regulars I'd rather not have to comb through AI spaghetti code to figure out the reasoning behind my ranking. AI loves complexity and it ends up being very inefficient.

The nickname and gendering is odd and I'd be careful getting too attached to it. It reads your tone and just tells you what you want to hear. I changed the axes on a few figures and fed it bad data, and proposed an entirely stupid theory about how tides work and it told me I was great and on the right track. We already had people believing dumb shit they read on Facebook, now we have a prompt that texts like they do and tell them they're on the right track.

Anything further will get this sent to the Politics section lol.

 

[Machine]

I specifically needed a dataset that contains one row for every season+game+player combination.

You can create this using an RDBS (SQL).

I have questions!

I am cautious of AI. I will use it to see how it would approach a solution, or to find basic syntax, but I would not trust it to do complex calculations. I would say a poker league could go either way. The AI should treat all the players the same, so if there was a bias, at least it should be applied evenly.

How do you handle players who joined mid-season in the 2nd or 3rd season?

I wonder if you're not trying to re-create the wheel. There is an ELO system used for Chess and/or multiplayer games. You might be able to find a system and workflow to plug N play.

Thanks for posting; it's sparked debate, and you can't be the first person who wanted to do something like this. I'm sure this has helped others.

 

[HiveKueen]

For my understanding, what ranking are you creating that is different than your points formula?

an all time ranking that spans all games played.

I specifically asked ChatGPT how I should calculate this given that the more games you play the harder it is to outperform because people learn how you play and adjust how they play against you. I've attached the formula we went back and forth to develop (after it initially suggested a Baysian average, which I rejected once it provided an explanation of it.)
I've attached a pic of the ranking formula.

Are you trying to create a ranking similar to that of how pro poker players are primarily ranked using the Global Poker Index (GPI)?

No. I was looking for a ranking system within the constraints of my league. Chatty compared the two and indicated that in her judgment the formula we developed together was more appropriate for my league.

How are you using these rankings?

I am not using them at all for anything. I was satisfying a curiosity. I have decided not to publish them to the league.

 

[HiveKueen]

I am cautious of AI. I will use it to see how it would approach a solution, or to find basic syntax, but I would not trust it to do complex calculations. I would say a poker league could go either way. The AI should treat all the players the same, so if there was a bias, at least it should be applied evenly.

I'm not sure I know what you mean here. could you give me an example of a bias in this context?

How do you handle players who joined mid-season in the 2nd or 3rd season?

I've attached the rankings formula to a different thread reply and a bit about how we came up with it.

I wonder if you're not trying to re-create the wheel. There is an ELO system used for Chess and/or multiplayer games. You might be able to find a system and workflow to plug N play.

I'll have to ask Chatty what ELO is and if there is a ranking system already out there that would work! haha

Thanks for posting; it's sparked debate, and you can't be the first person who wanted to do something like this. I'm sure this has helped others.

You're welcome.

 

[HiveKueen]

I use AI daily for work, but I think this type of system is most likely not going to be helpful. AI is an incredible tool , but one of the big problems now is that it'll say "SUCCESS! You did it! You're amazing!" when in reality it didn't do it, it chose to make up a bunch of stats it couldn't access instead. If I was one of your regulars I'd rather not have to comb through AI spaghetti code to figure out the reasoning behind my ranking.

I'm not going to publish the rankings. I do publish season standings and points-per-game info so players know where they stand for the purposes of qualifying for the season championships.

AI loves complexity and it ends up being very inefficient.

The nickname and gendering is odd and I'd be careful getting too attached to it. It reads your tone and just tells you what you want to hear.

I'll keep that in mind....

I changed the axes on a few figures and fed it bad data, and proposed an entirely stupid theory about how tides work and it told me I was great and on the right track. We already had people believing dumb shit they read on Facebook, now we have a prompt that texts like they do and tell them they're on the right track.

Anything further will get this sent to the Politics section lol.

 

[HiveKueen]

This does violate the TOS. It’s the first rule.

Be Polite​

• Do not post libelous, insulting, antagonistic, racist, discriminatory, aggressive, or otherwise abusive messages. If you have an issue with another member and cannot reply in a civilized and constructive way, you should ignore the user.

So do you want to self report or should we get one of the other regulars to do it?

Report yourself first! Your posts were simple rants and only served to undermine the value of mine!

 

[HiveKueen]

I spend about half my time doing data analysis and have found LLMs to speed things up a lot.

LLM???

 

[NotRealNameNoSir]

LLM???

Large Language Model. Look them up to understand what Chippy actually is.

 

[RadicusScout]

Would the World Golf Ranking be similar to what you're looking for?

They have a pretty detailed explanation here: https://www.owgr.com/how-the-ranking-works

 

[ekricket]

Report yourself first! Your posts were simple rants and only served to undermine the value of mine!

I didn’t insult you or attack you like you did me. I just made some comments on AI, didn’t even criticize anything you had done.

But since you started, it’s always the same. Call someone names or verbally insult them, then when they respond YOU are the victim.

All I did was comment on the use of AI and its trustworthiness. I see others did the same thing, expressing the same doubts, yet you cordially responded to them without insults.

Bias much?

 

[Dodger]

an all time ranking that spans all games played.

I specifically asked ChatGPT how I should calculate this given that the more games you play the harder it is to outperform because people learn how you play and adjust how they play against you. I've attached the formula we went back and forth to develop (after it initially suggested a Baysian average, which I rejected once it provided an explanation of it.)
I've attached a pic of the ranking formula.

No. I was looking for a ranking system within the constraints of my league. Chatty compared the two and indicated that in her judgment the formula we developed together was more appropriate for my league.

I am not using them at all for anything. I was satisfying a curiosity. I have decided not to publish them to the league.

I do find this interesting. I’m not so concerned with whether or not AI is being used to solve for this. I am intrigued by the ranking piece, and how it may differ from using league points.

If I may ramble…

Clearly, each league season has players that outperform others. Players earn points, and the points paint a picture of who is doing well and who is not. Do season standings tell you whether one player is better than another though? Maybe as a snapshot, but not over the long run. That seems to be what you are trying to answer here.

Point standings are objective and show a person’s position in a given season.

Rankings are subjective and evaluate players based on perceived strength and/or ability. Rankings also allow us to evaluate players who may not have played against each other in different seasons. Is the Season 7 champion a better player than the Season 8 champion?

That’s why I was asking in reference to GPI, as they seem to have this solved. As does the World Golf Ranking:

Would the World Golf Ranking be similar to what you're looking for?

They have a pretty detailed explanation here: https://www.owgr.com/how-the-ranking-works

Why can’t the GPI formula or World Golf Ranking formula be applied to a home league?

It sounds like you may have found a formula that works for your player rankings.

When looking at the results, do you tend to agree with what they show?

Like the College Football Playoff rankings, it’s time for the talking heads to argue every angle about which players should be above/below another.

 

[HiveKueen]

Would the World Golf Ranking be similar to what you're looking for?

They have a pretty detailed explanation here: https://www.owgr.com/how-the-ranking-works

Interesting, especially since, in a former life, I was a professional golfer who nearly made it to the big dance! I'll look into this - specifically , I will probably ask Chatty how my ranking and this ranking compare.

The thing that makes my ranking unique is that it takes my league's unique factors into consideration.

For example, the buy in is always $30, so, unlike the GPI for instance, the amount of the buy in isn't a factor in the ranking.

My ranking also factors in my unique points-per-game formula.

It factors in that we play poker specifically - so, for example it factors in that poker is a game of information and the more games yuo play against the same opponents the harder it gets to outperform them as they learn how you play and adjust to it.

It also adjusts for the fact that each player has played in a different number of games.

I haven't yet read about the golf ranking but I would be surprised if it didn't factor in the difficulty of the golf courses.

So, I kinda doubt I've recreated a wheel that is already readily available.

Cheers,
HK