Initial Ratings for USATT

[ttmonster]

Great that you can analyze your mistakes afterwards impersonally . I wish I could do that , I wish I could also do that when the game is going on! Isn't that what separates the men from the boys !

 

[Der_Echte]

Great that you can analyze your mistakes afterwards impersonally . I wish I could do that , I wish I could also do that when the game is going on! Isn't that what separates the men from the boys !

HAHAHA. you give me way too much credit for being non-emotional and cold calculating. If you were present for even a a match or two, you would likely not say that... well at least not until a few hours past by haha. I do not enjoy failing, but then again, not many do. I got better at a lot of things in life by failing A LOT so it will help later I hope. I had the potential to defeat everyone I lost to minus Asgaralli. I gotta get better. As much as I usually do not be hard-headed do the same thing, I did that a couple matches and it bit me right on my fat ass real hard.

 

[Der_Echte]

I like how they spelled your name. BTW that looks like the record of a 2050-2100 player. You did not lose to anyone under 2100 and you won vs 1 2100+ player and lost vs 2 2100+ players.

That is what it looked like that night, and the week prior. got 2nd each time. The botton end of that group are all 2000 playing lever irregardless of official rating, I see them at the club 3x a week. fred is an OX LP player who will attack at will if you play soft and is a countering monkey, Naveen is 2000 2 winger mean BH smooth joker, no way his playing level is under 1900.

I should be near dead last in that group in skill and playing level. Although it is possible to have a good night when some others have not their best night, there is no way I am even within one level of two of the guys I won vs. (Daniel and Hiep)

Maybe 5 years ago 2000 was a certain level, but right now in 2015/2016 it sure looks like a LOT of players hovering around 1900 are the same or better level than 2000 level players 5 years ago.

 

[Der_Echte]

I like how they spelled your name. BTW that looks like the record of a 2050-2100 player. You did not lose to anyone under 2100 and you won vs 1 2100+ player and lost vs 2 2100+ players.

On face value, if those were consistent results, then those would be the results of a 2000-2050 player on the rise.

In reality, all it really means is William played in some matches to the level he is capable of in two straight league sessions. My league record vs who i defeated:

Chopper Daniel: 1-1
Hiep: 1-0
Chopper Fred: 2-0
Naveen: 2-0

All it means is that I got a good start. ALL these guys are too good quality for me to hold such a winning percentage vs them. They will get used to me and the expected result will ensue. Perhaps playing these skilled dudes once a week will help me get better over time. Who knows? I certainly realize I need to get better after this weekend's tourney achieving only expected results and nothing special, even if it was a blast playing alone knowing next to no one there.

There is a local Baltimore player (Reggie) who I met in 2007 when I drove 1400 miles straight and went straight to Bogeyhunter's club. I was a 1400 level recreational player back then and needless to say, Reggie could easily defeat me. I saw Reggie after almost 10 years. In the warmup, I reminded him who I was and how sorry I looked trying to learn serves from Bogeyhunter in 2007 and utterly failed it looked so silly and then he remembered me. I told him I improved since then with some time in Korea. I said hello to him by taking game 1 of our U2000 group 11-2 enroute to a 3-0 win.

There was a chopper/retriever LP who simply stays alive in rallies and tries to get you to attack the first ball (big mistake, it always has something else on it or is differently placed) I didn't take the bait, took my time as he has no BH attack and only an occasion FH pick hit to keep you honest. I got by him 3-0 pretty easy doing what I wanted when I wanted, but I didn't open up on the first ball 90% of the time. The times I did attack, it was a power loop away from him or a very high spin opener and a smash or loopkill finish.

An hour after the match, he remembered me from my two visits to Baltimore in 2007 and told me I did not fall for the trap (like pretty much every one else does) and stayed patient and took the balls that were right.

So... there were a few positives, but those losses still burn. I earned each of my losses and deserved them. Eventually I will get to a level where I will not lose to those players again. Might have to wait until Asagalli get to be 90 yrs old though, that guy got calm and experience skills. Nothing shakes him.

 

[Der_Echte]

Since this thread was about erratic initial ratings and USATT not having a formula that matches reality to assign an initial rating...

When I got back from Korea and did a tourney in Hartford (Sep 2013 Open) this is my match record. Pretty much my top win was 2080 and my worst loss was 1913 (who defeated 2 players 2100 2200 level and next tourney finally got adjusted to 2100+) I came outta that shebang 1945. Realistically, that performance was over 2000 EASILY. Those is the breaks in USATT. 1900 today is 5 yrs ago 2000 level. I don't think it has ever been so tough to be a 2000 level player.



After a 2 year period with few tourneys and little practice and a bad injury, I recovered and did a tourney in DC area. The upper 1900s player I lost to Michael Levene is a 2200-2300 level former England #50ish. I easily defeated a 1920 chopper and a 3-1 win over a 1940s anti rubber player. I still barely made 1900 outta that shebang. Defeating those 1900s players like that was a 2000 level performance, but it doesn't work that way in USATT ratings.

 

[agold]

So this isn't completely on topic but here is the most accurate (I've found) rating estimator. It's usually been within 10 points for me, usually below the rating I actually get. http://www.breaking2000.com/rating-calculator/

 

[Der_Echte]

Abe is gunna break 2000 before the next president of USA is in office long enough to do the State of the Union speech.

Sent from my SM-N920V using Tapatalk

 

[Pnachtwey]

So this isn't completely on topic but here is the most accurate (I've found) rating estimator. It's usually been within 10 points for me, usually below the rating I actually get. http://www.breaking2000.com/rating-calculator/

I didn't see a link to download a ratings calculator. The only link I saw was for some pdf converter. Is the link gone now?

 

[Der_Echte]

I think he took down the calculator, because USATT came up with a new way or three to calculate ratings adjustments and initial rating. Now USATT is moving on to a new provider and we gotta see how that will work for us.

 

[Pnachtwey]

I have tweaked my evaluation routine a bit and improved it. There were flaws in the above method
It should be this.
pwl(r-1800)*pwl(2000-r)*pwl(2000-r)=???
The find the rating r that yields the highest probability ????.
It is more accurate than just averaging the best win or worst loss but the difference is only going to be about 40 points.
I can should a graph of probabilities if someone wants but no one seems to care if their ratings are off by 40 points or so.
A lot depends on if there is much difference in the ratings between the players. Wins against much lower and loses against much higher rating players have little effect on the new rating. 3 wins against 2000 players and 1 loss against a 2100 player would result in an initial rating of 2050 using the best win and worst loss method but if maximizing probabilities the initial rating for the new player would be 2099 which is 49 points higher. This is an extreme example because there are 3 wins against players with the same rating or have ratings that are close together. The best win/worst loss method ignores 2 of the wins.

It is easy to see if the probability curves are plotted out.

OK, I know. You don't care. Fine.

 

[Der_Echte]

Haha, Pnchy... jokers DO care a lot more than they let on, but count me in the crowd that doesn't care... well, at least doesn't obsessively care about it.

When players improve without 100 point kind of improvements, they do not get adjusted upwards before calculating the exchange of points.

This is a common situation with rapidly, players who do enough tourneys and avoid a major adjustment, unless it was a ridiculously fabulous tourney performance.

This is a natural deflationary pressure.

Look at my club's elite league players. Noveen at 1800s is WAY under-rated, he plays 200 level. Fred is even more of a ringer dangerous player 2000+ level. Daniel is easily over 2100 level. Hiep is around his level or 50 better skilled. Now suddenly I come around and have a few good wins, then let's say Noveen starts playing near 2100 level in a bit and Fred starts kickin' tail again... all at the expense of the top end of the elite, who are still the same 2200ish+ level players, yet now are under 2050 level due to 4 losses of 30+ points. Now everyone is correctly above 2000, but the elite 2200 level players are all under 2100, despite being the same 2200+ playing level.

That is deflationary pressure in action. That is what is happening to USATT the last 5-6 years. I'm not talking about teh juniors who are in a system that keeps them competing only vs other 200-400 level players who are all 1600+ by the time they move outta that group... I am talking about the top-down effect and hte bottom up effect. New better players ome in and are intially under-rated by formula, then as they win, they cause many existing elite players to lose a lot of points. Then, some fringe elite players get better and drop the lower elite players ratings... then some simply good players get better and drop the rating of that crowd...

 

[UpSideDownCarl]

The real trick would be to make it impossible for the sandbaggers to get away with sandbagging.

In tennis, if a guy is really a 6.5 (I think that is basically like 2550-2650) and he loses a bunch on purpose to get a 5.5 rating so he can enter easier events and simply win and take the money, they would see it, call the player on it and adjust his rating to where it should be and make him play in the correct bracket.

In table tennis, there are a bunch of guys who really should have a 2300 rating, who purposely drop matches to players who are around 1600-1800 to get their ratings to drop so they can enter easier brackets like the Under 2050 or Under 2150, and win the cash prize for 1st place.

This kind of thing makes it hard for the rating system to work properly.

There are guys who even do it where, in the round robin they just choose the weakest player to lose to and beat the other players in the round robin and either come in first or second in the round robin and still get to the elimination part of the tournament. So, the player can win the bracket (U2150), take the money from the cash prize, and still drop 50 points in rating.

No formula is going to work with those guys who are manipulating the system.


Sent from Deep Space by Abacus

 

[Der_Echte]

That takes the use of subjective judgment and someone(s) capable of that... not sure if we have enough common sense or willpower to do that. If we do crack down and get rid of ringers, they will find something else to be a ringer in. Look what happened when national assn everywhere blindly followed ITTF and made frictionless LP illegal for amatures... we lost a LOT of players and that will happen too, we already do not have enough players in the sport.

 

[Pnachtwey]

I just had to do the math and prove to myself there is a better way. I can look at equations, tables etc and see they aren't right but usually I must grind through the math before I have convinced myself they really aren't right. This is similar to the situation a few years ago when someone brought up handicap games. I can see the tables were wrong so I calculated my own but no one cares. It doesn't matter. To me it is a puzzle and challenge that must be solved.

The USATT rating system has too many fudge factors, phases and magic tricks to please me. I think the RC system is better.

I don't know what to do about ringers. That is a people problem and they are always more difficult to solve.

I also know that people can simply have a temporary case of ISuckAtThisItis or be on a hot streak. A ratings system will reflect the effect of both even thought that may not be the rating on the average.

I just HAD TO solve this problem. That is all.

 

[Loopadoop]

The USATT rating system has too many fudge factors, phases and magic tricks to please me. I think the RC system is better.

The USATT rating system is better or RC system would have been adopted by them.

A better system than either of them is the Class Rating System proposed at USA Hobby Table Tennis Coalition on Facebook.

 

[Suga D]

Generally i think these Rating systems are a good thing, cause they let you know where you're level's at the moment and can give you a good motivation for your training.
But anyhow i wouldn't expect a rating system to tell the full truth.

Cause there's not only ringers.

Sometimes when a player goes through some heavy changes in life (like a female player being pregnant, or losing a close family member, just like P. Baum did) his rating can drop if he doesn't play for a certain time. At least in the german rating system. Sometimes the rating might drop more than it should.
Or maybe someone who played on a high level and stopped before the rating system was existing and then returns to the sport and hence doesn't have a rating yet....

How do you want to calculate that?

So generally i wouldn't expect a rating system to be 100% accurate.
The rating score isn't everything.
I keep telling this to teammates who are facing higher ranked players, when i'm coaching them during a match.
And yes, sometimes they manage to beat a higher ranked player.

 

[pgpg]

The USATT rating system is better or RC system would have been adopted by them.

A better system than either of them is the Class Rating System proposed at USA Hobby Table Tennis Coalition on Facebook.

You do realize the irony of your post? Since class rating system is not widely adopted in US (your way too frequent mentions of it on every TT forum notwithstanding) it must follow that USATT ratings system is vastly superior!

 

[Pnachtwey]

The USATT rating system is better or RC system would have been adopted by them.

A better system than either of them is the Class Rating System proposed at USA Hobby Table Tennis Coalition on Facebook.

You haven't heard of not invented here syndrome and you can't believe how protective some people or organizations can be of their own work. I run into this all the time in industry too.

Go to this site and look at the very top. You will see the countries or leagues that use the RC system.
If you go to section 5 of this pdf you can see the smooth function ( a sigmoid function ) that the RC system an I used
http://www.ratingscentral.com/Doc/NewTTRS.pdf

I can give a few reasons why the RC system is better.
What you didn't seem to catch is that one of the first things I did was convert the USATT ratings tables to a smooth continuous function that is exactly like what the RC system uses. There are no artificial boundaries or ratings zones. The RC system does compute a uncertainty or standard deviation that is used to update the ratings instead of USATT fudge factors. The USATT rating system was designed for a time when computers weren't available.

Suga D. A rating system will reflect your performance and doesn't care why you played well or poorly. I think that is OK.

What is MUCH BIGGER PROBLEM and unforgivable are the directors that can't or don't implement their tournaments correctly resulting is erroneous ratings.

 

[Loopadoop]

You do realize the irony of your post? Since class rating system is not widely adopted in US (your way too frequent mentions of it on every TT forum notwithstanding) it must follow that USATT ratings system is vastly superior!

By usage, the USATT rating system, is superior in the USA.

By Concept, a Class System, appears to be superior to a Match System. That is Debatable.

 

[Loopadoop]

You haven't heard of not invented here syndrome and you can't believe how protective some people or organizations can be of their own work. I run into this all the time in industry too.

Go to this site and look at the very top. You will see the countries or leagues that use the RC system.
If you go to section 5 of this pdf you can see the smooth function ( a sigmoid function ) that the RC system an I used
http://www.ratingscentral.com/Doc/NewTTRS.pdf

I can give a few reasons why the RC system is better.
What you didn't seem to catch is that one of the first things I did was convert the USATT ratings tables to a smooth continuous function that is exactly like what the RC system uses. There are no artificial boundaries or ratings zones. The RC system does compute a uncertainty or standard deviation that is used to update the ratings instead of USATT fudge factors. The USATT rating system was designed for a time when computers weren't available.

Suga D. A rating system will reflect your performance and doesn't care why you played well or poorly. I think that is OK.

What is MUCH BIGGER PROBLEM and unforgivable are the directors that can't or don't implement their tournaments correctly resulting is erroneous ratings.

The USATT had a chance to adopt the RC System many years ago but chose not to after their board of directors studied both systems.

The RC System appears to work good when one class of players like professionals are grouped together. When you have a wide range skill level of players like amateurs who play on an inconsistent basis, there are wild fluctuations in calculations, so it appears that it doesn't work very good.