Web page that calculates APA 8-ball rating from win/loss record

[Skippy27]

That's only if you consider the game on a per-shot basis.

When I was a 4, I would hit in every ball I could. I often ended up with a problem ball or two that I had to bat around the table for several innings and I would usually lose my games.

I quickly learned that trying as hard as I could to make every ball I could was a losing strategy. As soon as I figured that out I moved up to a 5 and now I usually win my games/matches.

So I adopted this style of play so I could win games and matches, which is the opposite of sandbagging.

Not trying 100% on every single one of my shots is strategically better for me, so I'm actually playing as well as I can.

Remember, pool is about winning the game, not sinking balls.

I will assume by not trying 100% you mean not trying to pocket a ball. FYI, that is not sandbagging. You are playing a good strategic game and playing defenses/safeties and they should be marked accordingly. Sandbagging would be while doing the exact same thing you speak of and then trying to disguise or argue whether or not they were defensive shots.

Good players, not sandbagging, will make the balls they can and try to get breakouts that they need. If they fail the breakouts and/or see no out for the game they will play defensively as they should as as you say you did above. That is perfectly fine.

Let me continue on with my example above as this was discussed last night at the scoring clinic we had that was put on by actual APA employees being the league owns and operates our area right now. The above 4 I mentioned that should be a 5, missed 3 attempts at an 8 ball that was about 6 inches from the corner. Straight up missed it 3 tries in a row while she was playing a 2, that would literally be a 1 or less if they had those levels. The 6 that should be a 7 was playing a 3 (I believe) and scratched probably more times then made balls and lost to this 3.

The 7 on our team spoke with the officials there and their only response was to be sure to mark the shots as safeties, then the innings mean nothing and they can do it 100 times and they can lose every match for the next 5 years and it will have no bearing on their rating. That is straight from the APA. I also liked to hear that they have people from other areas come in and watch players covertly.

 

[tomker]

Well you can't assume that, so you haven't proven anything. Furthermore, 400 ELO points isn't standard, 400 ELO points is different in every ELO system, because of varying k-factors among other variables. If you want to try to prove APA matches could be more fair, then please stop comparing it to ELO-based systems, and do analysis based solely on the APA's system. You're already biased because you don't believe in the innings-per-win metric.
...

Let's examine this situation. Elo ratings are simply numbers that indicate how likely one player is to win a game against another player. The higher the number, the more likely the player is to win, and thus the better the player is.

Thus, even if APA skill levels are calculated in a different way, they must correspond to ranges of Elo ratings.

Otherwise what you're saying is that higher rated APA players aren't necessarily expected to win vs. lower rated APA players... that a 6 isn't necessarily more likely to win games against a 5... and if that's the case, why bother pretending to have skill levels at all?

 

[Don Owen]

My partner and I have been studying rating systems. We have access to a database of matches played between 2009 and 2014 containing approximately 9,000 players, 33,000 matches and 250,000 games. We hired Dr. Mark Glickman, developer of the Glicko and Glicko-2 rating systems, used extensively in Chess, to help us.
Dr. Glickman did a very interesting thing. He first rated the players using the first 2/3 of the matches played and then used those ratings to see how well the ratings predicted the last 1/3 of the matches. We were disappointed. That disappointment led us to seek a different approach.
I suggest that all rating systems should be tested that way.
Stay tuned,
Don

 

[tomker]

...
If you want to waste your time to find more fair matches that fit within the current time constraints, go ahead. It really boggles my mind how arrogant you must be to think that you are smarter than the biggest most successful league there ever was. Like the APA doesn't have any math majors working for them and they haven't invested millions in coming up with their system. What isn't getting through your thick skull is that there are so many other factors than making matches perfectly fair that factor into the equation.

Clearly, they are doing something right, otherwise there would have been plenty other people like you offering fair matches for all players that would have taken over the APA years ago.

You've made this argument a couple of times, that the APA must be good because it's big. That's like arguing that McDonalds food is healthy because a lot of people eat it, like it, and pay for it.

The APA rating system is broken. Every higher-level tournament I attend, I'm surrounded by people who are constantly talking about manipulating their ratings or talking about who's overrated and who's underrated. There are systems set up to have secret committees review peoples records and play because they have to manually change peoples' ratings. There are dozens of "observers" in Vegas roaming around watching people play to do the same. Maybe you just assume this sort of thing is normal and necessary but it isn't. I've been to chess tournaments and this doesn't happen because their rating system is open, everybody understands it, and there are well-understood and generally-accepted measures that prevent/deter manipulation.

You make a lot of complaints about me being new to the community and being arrogant to think that my ideas are somehow better than the APA's. Couldn't be further from the truth, because "my" ideas are not my ideas. I didn't invent the idea that "security through obscurity" is bad and I didn't invent the idea of random variables to indicate skill, which can be used for rating and handicapping and so forth. In fact, the situation is kind of the opposite. The idea of you simply assuming that the APA's system is better than "my" ideas is arrogant, since "my" ideas have been around much longer than the APA, are accepted by millions, have been used to divide up probably hundreds of millions in prize money, and have have been peer-reviewed countless times (unlike the APA's secret system).

But either way, I don't know how I got into a discussion of all of this. The point of the thread was to post a link to a program that I thought was interesting. I guess the problem is that people started posting seemingly-ridiculous claims, e.g., that the APA handicapping table isn't even intended to be fair. Whatever, even if that's true, I don't really care and I'm still going to play in the APA anyway because I think it's fun.

 

[railbird99]

Let's examine this situation. Elo ratings are simply numbers that indicate how likely one player is to win a game against another player. The higher the number, the more likely the player is to win, and thus the better the player is.

Thus, even if APA skill levels are calculated in a different way, they must correspond to ranges of Elo ratings.

Otherwise what you're saying is that higher rated APA players aren't necessarily expected to win vs. lower rated APA players... that a 6 isn't necessarily more likely to win games against a 5... and if that's the case, why bother pretending to have skill levels at all?

Are you deliberately being dense? The basis for your calculations relied on the assumption that skill levels were separated by a range of 400 ELO points. It's this number 400 that you came up with that is not only wrong, but when you say 400 ELO, it has no meaning whatsoever because it is a relative measurement. Every ELO system is different, with different k-factors and other variables, so 400 in one system means something entirely different in another system.

As far as being able to translate APA ratings into a range of ELO numbers in general, that is quite obviously false. APA ratings are based on innings per win, while ELO ratings are based on wins/losses, so there is no direct conversion. Even if you could say SL 6' are between 400 ELO and 500ELO, it doesn't mean that 100ELO range is constant for other skill levels. Not because the skill level difference varies, but because the underlying mechanism for determining skill level is different between ELO and the APA system.

 

[tomker]

Are you deliberately being dense? The basis for your calculations relied on the assumption that skill levels were separated by a range of 400 ELO points. It's this number 400 that you came up with that is not only wrong, but when you say 400 ELO, it has no meaning whatsoever because it is a relative measurement. Every ELO system is different, with different k-factors and other variables, so 400 in one system means something entirely different in another system.

As far as being able to translate APA ratings into a range of ELO numbers in general, that is quite obviously false. APA ratings are based on innings per win, while ELO ratings are based on wins/losses, so there is no direct conversion. Even if you could say SL 6' are between 400 ELO and 500ELO, it doesn't mean that 100ELO range is constant for other skill levels. Not because the skill level difference varies, but because the underlying mechanism for determining skill level is different between ELO and the APA system.

I didn't say anything about the 400 point range because you didn't offer any other range that you wanted me to use. You're right that it doesn't mean anything by itself but since the number is 400 I figured you would recognize it as the typical number used to calculate win/loss probabilities in chess. Look at this Wikipedia page and search for "algorithm of 400":

http://en.wikipedia.org/wiki/Elo_rating_system

So no, I'm not trying to be dense and I don't appreciate the tone. In fact I could accuse you of being purposely dense by not recognizing this 400 situation but I think it's just becoming more and more clear that you are not very familiar with rating systems and how they're designed and implemented. And in that case I'd appreciate a little more rhetorical respect.

As for APA ratings not corresponding to Elo ratings, again, what you are literally saying is that higher rated APA players are not necessarily expected to win more games against lower rated players. And if that's not the case, again, what's the point of having skill levels?

That's actually my point with all of this nonsense about innings-to-win. Ultimately you want your rating numbers to indicate who's more likely to win, so if you can't even connect the dots from rating numbers to win probabilities then you almost might as well base your rating numbers on who has the most expensive cue.

 

[railbird99]

You've made this argument a couple of times, that the APA must be good because it's big. That's like arguing that McDonalds food is healthy because a lot of people eat it, like it, and pay for it.

The APA rating system is broken. Every higher-level tournament I attend, I'm surrounded by people who are constantly talking about manipulating their ratings or talking about who's overrated and who's underrated. There are systems set up to have secret committees review peoples records and play because they have to manually change peoples' ratings. There are dozens of "observers" in Vegas roaming around watching people play to do the same. Maybe you just assume this sort of thing is normal and necessary but it isn't. I've been to chess tournaments and this doesn't happen because their rating system is open, everybody understands it, and there are well-understood and generally-accepted measures that prevent/deter manipulation.

You make a lot of complaints about me being new to the community and being arrogant to think that my ideas are somehow better than the APA's. Couldn't be further from the truth, because "my" ideas are not my ideas. I didn't invent the idea that "security through obscurity" is bad and I didn't invent the idea of random variables to indicate skill, which can be used for rating and handicapping and so forth. In fact, the situation is kind of the opposite. The idea of you simply assuming that the APA's system is better than "my" ideas is arrogant, since "my" ideas have been around much longer than the APA, are accepted by millions, have been used to divide up probably hundreds of millions in prize money, and have have been peer-reviewed countless times (unlike the APA's secret system).

But either way, I don't know how I got into a discussion of all of this. The point of the thread was to post a link to a program that I thought was interesting. I guess the problem is that people started posting seemingly-ridiculous claims, e.g., that the APA handicapping table isn't even intended to be fair. Whatever, even if that's true, I don't really care and I'm still going to play in the APA anyway because I think it's fun.

You are arrogant and not really clever because you are taking the same principles that everyone is fully aware of and trying to apply them to a system that already works great for its intended purpose. You are trying to fit a square peg in a round hole.

I'm not arguing that the system you propose is bad. In fact I'm completely agreeing that an ELO system would be great. I'm going to write this in all caps so I can be clear for the 100th time:

IT IS IMPOSSIBLE TO ACHIEVE FAIR MATCHES WHEN HANDICAPPING 8-BALL BY GAMES, NO MATTER HOW ACCURATE THE MECHANISM IS THAT DETERMINES SKILL LEVEL. IT WILL NOT TRANSLATE PERFECTLY INTO THE SHORT MATCHES NEEDED GIVEN THE TIME CONSTRAINTS. ALL SYSTEMS ARE FLAWED.

You are taking a 101 statistics class education and assuming you now have the best possible system in place. That's what makes you arrogant. I'm not saying either system is better than the other for handicapping 8-ball matches, just that the APA's system works perfectly well for its intended purpose.

For the hundredth time, matches being fair is not the only factor that the APA has to consider when devising their system. You are unable to grasp the full scope of the issue, and you are stuck in your little bubble thinking solving the probability of someone winning a match thus solves the entire issue.

For the hundredth time, this isn't security through obscurity. It happens to be hidden, just like the source code for microsoft windows, that doesn't mean it relies on the fact that it's hidden for its effectiveness.

This is getting repetitive and old. Take your high school math somewhere else and start your own league. You've clearly got it all figured out.

 

[tomker]

Good players, not sandbagging, will make the balls they can and try to get breakouts that they need. If they fail the breakouts and/or see no out for the game they will play defensively as they should as as you say you did above. That is perfectly fine.

What everybody is saying is makes a lot of sense. If I took the game more seriously and was more thoughtful about it, then all of my shots would either be trying to pot a ball or playing a D.

The fact is that I play pool for fun. I drink beer, hang out with my friends, and try to win at pool. I don't have the mental stamina or the desire to treat every shot like a life-or-death situation, which is what would be necessary for me to always be playing to 100% of my ability. There are a lot of times where I can't decide if I really want a ball to go in or not, so I just take a swing at it and see what happens. I know this means I'm obviously not a very good player, and I'm not proud of that, but I don't think I'm sandbagging because I'm still trying to win the game and I have no desire to manipulate the rating system.

 

[railbird99]

I didn't say anything about the 400 point range because you didn't offer any other range that you wanted me to use. You're right that it doesn't mean anything by itself but since the number is 400 I figured you would recognize it as the typical number used to calculate win/loss probabilities in chess. Look at this Wikipedia page and search for "algorithm of 400":

http://en.wikipedia.org/wiki/Elo_rating_system

So no, I'm not trying to be dense and I don't appreciate the tone. In fact I could accuse you of being purposely dense by not recognizing this 400 situation but I think it's just becoming more and more clear that you are not very familiar with rating systems and how they're designed and implemented. And in that case I'd appreciate a little more rhetorical respect.

As for APA ratings not corresponding to Elo ratings, again, what you are literally saying is that higher rated APA players are not necessarily expected to win more games against lower rated players. And if that's not the case, again, what's the point of having skill levels?

That's actually my point with all of this nonsense about innings-to-win. Ultimately you want your rating numbers to indicate who's more likely to win, so if you can't even connect the dots from rating numbers to win probabilities then you almost might as well base your rating numbers on who has the most expensive cue.

You are completely missing the point. The APA IS NOT AN ELO SYSTEM. The number 400 in this context is ridiculous. Stop googling stuff you know nothing about.

ELO is a great system, but stop perpetuating it as the almighty determinant of actual skill level and the percentage chance that someone will win a match. The ELO system attempts to ESTIMATE the percentage chance each player will win the game. The estimation becomes more and more accurate the more games that are played, but it never reaches 100% accuracy.

So for the last time, you can not base your calculations for how accurate APA matches are on an arbitrary rating in an unrelated system.

 

[tomker]

...
This is getting repetitive and old. Take your high school math somewhere else and start your own league. You've clearly got it all figured out.

The reason it's getting so repetitive and old for you is because you're not saying anything to my point. I'm saying the races could be more fair and all you're saying is that there's some inherent level of unfairness.

Can't you see that both might be true? And in fact I have always agreed with you that there's some inherent level of unfairness. So now you are getting yourself worked up and blue in the face repeating yourself over and over when I'VE BEEN AGREEING WITH YOU THIS WHOLE TIME.

Yet this entire time you have not said anything to my point that the matches could be more fair. If I made a concrete statement saying that I think the 4 vs. 6 race could be more fair by switching it to 3-6 instead of 2-5, what would you say about that particular case? Because this is exactly the sort of argument I've been making this entire time.

 

[tomker]

...
ELO is a great system, but stop perpetuating it as the almighty determinant of actual skill level and the percentage chance that someone will win a match. The ELO system attempts to ESTIMATE the percentage chance each player will win the game. The estimation becomes more and more accurate the more games that are played, but it never reaches 100% accuracy.

So for the last time, you can not base your calculations for how accurate APA matches are on an arbitrary rating in an unrelated system.

You are conflating the numbers (which are just indications of probability) with the "system" of how the numbers are updated.

Higher number = better.

Same as APA. If you want to call it the "higher number better" system so you don't get confused with all the other emotional baggage about the Elo system then that's fine with me. You can do a search-and-replace on my posts.

 

[railbird99]

What everybody is saying is makes a lot of sense. If I took the game more seriously and was more thoughtful about it, then all of my shots would either be trying to pot a ball or playing a D.

The fact is that I play pool for fun. I drink beer, hang out with my friends, and try to win at pool. I don't have the mental stamina or the desire to treat every shot like a life-or-death situation, which is what would be necessary for me to always be playing to 100% of my ability. There are a lot of times where I can't decide if I really want a ball to go in or not, so I just take a swing at it and see what happens. I know this means I'm obviously not a very good player, and I'm not proud of that, but I don't think I'm sandbagging because I'm still trying to win the game and I have no desire to manipulate the rating system.

Dude seriously, what is your deal.

Noone is saying you should ever act like it's life or death, because it is just for fun. You are attempting to make your point sound more reasonable by taking what we are saying to unreasonable extremes.

The point is you should never purposefully "not try" and "not care" if the shot goes in and not even make a choice as to whether you are trying to make it or miss it, period. Some shots you won't care as much if it goes in, because you are also playing safe at the same time. That's completely different than intentionally "lazily" shooting it without at least committing to either missing it or making it.

 

[tomker]

My partner and I have been studying rating systems. We have access to a database of matches played between 2009 and 2014 containing approximately 9,000 players, 33,000 matches and 250,000 games. We hired Dr. Mark Glickman, developer of the Glicko and Glicko-2 rating systems, used extensively in Chess, to help us.
Dr. Glickman did a very interesting thing. He first rated the players using the first 2/3 of the matches played and then used those ratings to see how well the ratings predicted the last 1/3 of the matches. We were disappointed. That disappointment led us to seek a different approach.
I suggest that all rating systems should be tested that way.
Stay tuned,
Don

This is brilliant news. I would love to know more about your project.

How did you get the data?

If it's public data, you could run a Netflix-like contest:

http://en.wikipedia.org/wiki/Netflix_Prize

I mean, obviously don't put up a million dollars in prize money but I'm sure there are many people out there who would like nothing better than to try to come up with a good rating system against this data. Myself included.

I remember spending several happy weeks working on the Netflix data.

 

[tomker]

Dude seriously, what is your deal.

Noone is saying you should ever act like it's life or death, because it is just for fun. You are attempting to make your point sound more reasonable by taking what we are saying to unreasonable extremes.
...

It's called proof by induction.

If everybody is cool with only trying 95% on some shots and that's not sandbagging, then why is it sandbagging if I only try 50% on some shots.

What percentage do I have to hit so nobody accuses me of sandbagging?

 

[railbird99]

You are conflating the numbers (which are just indications of probability) with the "system" of how the numbers are updated.

Higher number = better.

Same as APA. If you want to call it the "higher number better" system so you don't get confused with all the other emotional baggage about the Elo system then that's fine with me. You can do a search-and-replace on my posts.

I feel like I'm in grade school again. Yes billy, higher means better. No billy, the relative difference in ELO ratings is not the same as the relative difference in innings per win. So no billy, you can't use 400 ELO points as the constant range for skill levels when trying to do any credible calculations regarding the APA's system. And billy, ELO is a more complex calculation that "attempts" to accurately turn wins/losses into a single number metric, therefore ELO is not a direct representation of wins/losses.

 

[tomker]

I feel like I'm in grade school again. Yes billy, higher means better. No billy, the relative difference in ELO ratings is not the same as the relative difference in innings per win. So no billy, you can't use 400 ELO points as the constant range for skill levels when trying to do any credible calculations regarding the APA's system.

Okay, I wrote a program to do this, so all you have to do is tell me which ELO numbers you want me to plug in for the corresponding APA numbers and I can tell you how fair each race is.

 

[railbird99]

It's called proof by induction.

If everybody is cool with only trying 95% on some shots and that's not sandbagging, then why is it sandbagging if I only try 50% on some shots.

What percentage do I have to hit so nobody accuses me of sandbagging?

The point is you are playing two different shots without committing to either one of them. It's not about how much you try. With regards to how much you try once you have made a decision to miss or make the shot, you shouldn't be purposefully trying less, meaning most people don't consciously try less. There are other factors such as mood, how good you are playing, etc. People don't make conscious decisions like, "well this shot could go either way so I'm going to try 50% here."

Furthermore, how is any handicap system supposed to be accurate when you are varying how much effort you are putting into each shot?

Stop being ridiculous please.

 

[railbird99]

Okay, I wrote a program to do this, so all you have to do is tell me which ELO numbers you want me to plug in for the corresponding APA numbers and I can tell you how fair each race is.

It's the fact that an arbitrary number will change your results that invalidates your calculations

 

[tomker]

The point is you are playing two different shots without committing to either one of them. It's not about how much you try. With regards to how much you try once you have made a decision to miss or make the shot, you shouldn't be purposefully trying less, meaning most people don't consciously try less. There are other factors such as mood, how good you are playing, etc. People don't make conscious decisions like, "well this shot could go either way so I'm going to try 50% here."

Furthermore, how is any handicap system supposed to be accurate when you are varying how much effort you are putting into each shot?

Stop being ridiculous please.

Sorry, it's what I do. I'm not a skilled enough pool player that I always know what to do and if I think the outcome of a shot is likely 50-50 to improve my situation then I will simply not try very hard on that shot.

I don't think this is unreasonable and so far nobody has made a coherent argument for this being sandbagging, since I'm still trying to win the game.

 

[tomker]

It's the fact that an arbitrary number will change your results that invalidates your calculations

Okay, well, I tried to move the conversation forward by doing some mathematical analysis.

Of course I had to make certain assumptions because the APA rating system is secret.

If you're not going to offer any concrete arguments about anything, I think we're done here?