|
|
| |||||||
| From | Message | ||||||
|---|---|---|---|---|---|---|---|
|
 Team Game Ratingteam game rating as the primary statistic when making pairings. Yes, this is radical, but it makes a lot of sense if you think about it. How many players have you seen that have a higher team game rating than current rating? It is very rare and a player to avoid. Look at your own team game rating, it is anywhere from 100-400 points lower than your current rating. The higher the current rating the larger difference in team game rating. There is always a degree of resistance to change but if you desire fair team games, using the team game rating will do that. I no longer make our team's pairings. George bibio5 is the primary match maker for Kings of the Castle now. |
||||||
|
The problem I envisage is getting all teams to agree. |
||||||
|
to resist change but eventually change wins out. I started this thread hoping other Captains would at least consider this. How can a player manipulate a team game rating? |
||||||
|
Algebraic Notation Now we have FEN notation, which is fun. Anyone ask GK to put the team rating on the player pop up window? There are a few other improvements long ignored. I wonder if the new owners retained any programmers, or dismissed them? I imagine this was mostly one guy working from his mother’s basement. |
||||||
|
listen more closely to our ideas and act when appropriate.. |
||||||
|
No doubt it is easier with algebraic notation. |
||||||
|
Regarding GameKnot's improving the site - no way! They're just cruising along collecting our subscription fees & doing nothing else. |
||||||
|
|
||||||
|
Basically, pair as normal +/- 50 rating points, but using the team game ratings. (my preference would be +/- 25). 90 day high rating would be ignored. Sure, some current ratings matches could be more than +/- 50 current rating. Captains can still decline challenges of course. There never was a rule on how to pair team players. |
||||||
|
One of the three players mentioned was paradiesvogel, current rating 1520. His team game rating is actually higher than his current rating (1533). If paired against a player with equal team game rating the current rating would be much higher, 1700's. This is why rating manipulation would disappear. |
||||||
|
paradiesvogel would drop like a rock! |
||||||
|
I Didn't Used to Know It is a current rating but just for team games. A player who purposely loses non-team games but puts full effort into team games will have an accurate team rating and an artificially deflated main rating. A player who puts full effort into all of his games will have an accurate team rating and an accurate main rating. By "accurate," I mean that the rating will be what the player has earned based upon recent results. An "accurate" rating could still exaggerate a player's strength or underestimate a player's strength if that player has been on a recent hot streak or slump. |
||||||
|
the fairness of the match. Hopefully, the matches would pair individuals with like stats. |
||||||
|
into team games will have an accurate team rating and an artificially deflated main rating.". What I mean is that "accurate team rating" is inflated because his opponents were underrated. Just look at paradiesvogel's stats. |
||||||
|
This Is Like Edit #8 of My Attempts to Explain What I Mean without Confusing Myself Greatly |
||||||
|
Mikemate is Right |
||||||
|
Isn't the point of pairing by team rating that the team rating is the rating that is accurate? I agree that Mario's win ratio in team games will rise. I agree that his score in team games will rise. But his team rating will be much higher than the team rating for the children's box car derbies. So Mario will gain very few points in team rating for each team win but will lose many points for the rare loss (or draw!) that occurs when his car stalls or he gets a flat tire during a team game. If the Elo rating system is working as it should, then Mario's team rating will be correct. Mario's main rating, on the other hand, will be low enough from all the bogus losses in non-team games to make it appear from that rating that children's box car derbies are appropriate opponents for him. |
||||||
|
I consulted Wikipedia for driver names But if A.J. Foyt, like most racers but unlike Mario, is putting a full effort into his non-team races, then Foyt's main rating will be much higher than Mario's. Also, A.J. Foyt, who is busy racing versus racers such as Jim Clark and Phil Hill in team races, will have a much lower team score and much lower team win ratio than Mario, who is racing children's box car derbies in team races. Mario's extraordinary team score and his team rating that is out of line with the team ratings for his children's box car derby peers are clues that something fishy is going on. If captains paid attention to how extraordinary Mario's team winning percentage is compared to his overall winning percentage, then he would have a lot more difficulty getting team matches based upon his bogus main rating. |
||||||
|
Anyhow... |
||||||
|
If team captains are using only team ratings for matches and ignoring the current rating as you stipulate, then yes—that is Mike’s point also, I believe. I thought you were disagreeing. |
||||||
|
I think team rating is a valid method for pairing |
||||||
|
|
||||||
|
chrispall (1399) vs replicannt (1377) ozdin (1260) vs fishist (1249) Look at their team game ratings. It can be done! |
||||||
|
Using team ratings works but 75% of the teams refuse to use this, so good luck with this. |
||||||
|
1. Current average: +-50 either way; 2. 90-day average: +-50 either way; 3. one yer average: +-100 either way; 4. past game history 5. 90-day high: +-100 either way. If this is inconclusive, the all time average is considered, and then the team average. You can take note that the team average is of a low priority for team SWAT. Statistics such as the highest player won means little, as this could be due to a number of issues, such as a time-out, a huge stroke of luck, not skill, sandbagging (we have seen this) or a player who leaves the site for any reason. I would consider a player with less than 200 games a new player, whose statistics have not settled yet, and I would be wary of such a player. I watch time-outs as well; if it is a significant percentage, it could be a double edged sword: my player has a good chance to win based on a time-out, or the opponent is deliberately sandbagging and is actually much stronger than my player. One should be very careful of a high time-out percentage. It is virtually impossible to find the last string of time-outs with the search options we have available, so one cannot judge if it is at a specific time of year, a once-off event or a continuous trend. However, a factor that I consider extremely valuable is trust. Just this morning I sent a message to orkneylad, confessing that my player had a string of 19 timeouts (due to ill health) and the current average as well as the 90-day average must not be underestimated. My player can reasonably be expected to perform well above his current statistical pointers. I do expect the same honesty from opposing captains, but the truth is that I only trust three other captains to be so honest and transparent. There are two captains in particular that I won't trust with anything, and I double check everything they send or answer with. Both have a proven record (with team SWAT at least) to replace fair nominations with players who are massively overwhelming in the contest- according to our priorities of the statistics. I have always maintained that any contest must be fair both ways. If I do nominate an unfair player, I always provide an explanation, and of course an invitation to decline without prejudice. If we don't set up fair matches our players will leave the teams and either play for themselves, or join a team where they can count on their captain to protect them from unfair matches. |
||||||
|
And I totally avoid that one team who always makes unfair substitutions; guess I've been lucky & missed the other. |
||||||
|
I'm fairly sure that team game rating will never be used as the primary preference for making matches. However, using current rating and/or current average, 90-day average, and one year average, will NOT protect your mate from being paired with a rating manipulator. Pairing by team game rating is the only way to avoid them. How can a player manipulate his/her team game rating (very minimal). I no longer seek matches. Gameknot could make the team experience much better, we all know this. The number of teams on Gameknot has steadily declined ever since I was here. Why? Maybe it is the way and time it takes to make matches. Gameknot could and should streamline this. If I started making matches today I would use the team game rating almost exclusively. |
||||||
|
So Many Timeouts... The rating graphs can give an indication of when the timeouts occurred if there have been a lot of them. But, yeah, if a captain really wants to know, there is nothing to do but go through the player's game history and click on games and look for the timeouts that way. Batches of consecutive losses provide a starting point. Trying to locate the timeouts is tedious. Timeouts are one of the more disheartening parts of GameKnot captaining for me. They also can totally distort all of a player's statistics, including the player's team rating. Often there is a very valid explanation for a player's timeouts. But often there is not. In many instances the explanation is that the player is not very committed to his or her games on GameKnot. <<riaannieman: "I watch time-outs as well; if it is a significant percentage, it could be a double edged sword: my player has a good chance to win based on a time-out, or the opponent is deliberately sandbagging and is actually much stronger than my player.">> One of my least favorite types of matches that I sometimes agree to anyway is one in which two players are a current-rating "match" only because one of the players has a large number of recent timeouts. In such instances it sometimes occurs to me that the match may indeed be approximately equal, but the *way* that it is equal goes something like this: The opposing player wins 50%; my player wins by checkmate or resignation 30%; my player wins by timeout 20%. Or sometimes my player is the one with the timeouts. It's not as if my team is immune to them either. |
||||||
|
Result Column Be easy to glance through a few pages of games, or sort the timeout losses to see the game dates this way. |
||||||
|
|
||||||
| |||||||
