Author |
Topic: 2010 World Championship (Read 16510 times) |
|
Adanac
Forum Guru
Arimaa player #892
Gender:
Posts: 635
|
|
Re: 2010 World Championship
« Reply #60 on: Jan 25th, 2010, 2:52pm » |
Quote Modify
|
I’m happy that the scheduler worked for round 4. Apparently, though, I still have misconceptions about how the process works. When I look at the players with 2-1 records they are: (3) Tuks (4) Fritzlein (5) The_Jeh (6) Nevermind (7) Nombril (8 ) 99of9 Normally the pairings would be 3-6, 4-7, 5-8 but due to the no-repeat-matchup rule it checked 3-6, 3-7, 3-8 until it found a legal pairing. So far, so good: (3) Tuks vs. (8 ) 99of9 Next I expected 4-6 and 5-7 which are both legal matchups: (4) Fritzlein vs. (6) Nevermind (5) The_Jeh vs. (7) Nombril Instead, it kept the 4-7 pairing from the original assumptions and then filled in 5-6 as the final pairing. So it’s a good thing I added a bold disclaimer to my previous post because Nombril and Nevermind now have much different pairings than I anticipated.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #61 on: Jan 26th, 2010, 11:23am » |
Quote Modify
|
The new strength of schedule kicks in for the standings after round 3, and it is having the desired effect. For the 1-2 players, naveed's opponents have 3, 1, and 0 wins, while omar and Hippo each faced opponents of 2, 1, and 1 wins. In the old SoS, those would be considered equally tough schedules, but the new SoS considers naveed's schedule weaker. Intuitively it is because naveed faced one stronger, one equal, and one weaker opponent, whereas omar and Hippo each faced one stronger, two equal, and no weaker opponents. The classification of stronger/equal/weaker is more relevant than how much stronger or weaker. By the same token, 99of9 at 2-1 has faced a weaker schedule than Nombril, The_Jeh, and Nevermid who are also at 2-1. The latter three each faced opponents of 2, 1, and 0 wins, while 99of9 faced opponents of 1, 1, and 1 win. In the old SoS they all would be tied at 3 points, but the new SoS recognizes that 99of9 faced three weaker opponents, whereas the others each faced one equal and two weaker opponents. So far the new SoS has only broken ties that the old SoS would have left intact. The new SoS has not reversed any rankings where the old SoS would have put one player higher than another. I expect, however, that in the final tournament standings there will be at least one reversal between the two SoS calculations. It will be a true test of the new SoS to see whether we like what it does in the case where it makes the biggest difference.
|
|
IP Logged |
|
|
|
woh
Forum Guru
Arimaa player #2128
Gender:
Posts: 254
|
|
Re: 2010 World Championship
« Reply #62 on: Jan 26th, 2010, 3:55pm » |
Quote Modify
|
on Jan 26th, 2010, 11:23am, Fritzlein wrote:By the same token, 99of9 at 2-1 has faced a weaker schedule than Nombril, The_Jeh, and Nevermid who are also at 2-1. The latter three each faced opponents of 2, 1, and 0 wins, while 99of9 faced opponents of 1, 1, and 1 win. In the old SoS they all would be tied at 3 points, but the new SoS recognizes that 99of9 faced three weaker opponents, whereas the others each faced one equal and two weaker opponents. |
| WHR however, puts 99of9 between those 3 players, just ahead of The_Jeh. 1. Adanac 1763 2. chessandgo 1742 3. Tuks 1648 4. Fritzlein 1609 5. Nombril 1561 6. Nevermind 1554 7. 99of9 1535 8. The_Jeh 1530 9. Simon 1493 10. PMertens 1461 11. woh 1422 12. Hippo 1406 13. omar 1400 14. naveed 1395 15. ChrisB 1255 16. fritlforpresident 1225 Looking at their opponents ranking 99of9 The_Jeh 1493 1609 ~ -120 1400 1406 ~ = 1422 1225 ~ +200 Overall the opponents of 99of9 are a bit stronger.
|
« Last Edit: Jan 26th, 2010, 4:21pm by woh » |
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #63 on: Jan 26th, 2010, 4:18pm » |
Quote Modify
|
One advantage of using in-tournament WHR for the SoS is that there are already no ties after three rounds! I didn't think of this ahead of time. My new SoS has fewer ties than the old one, but there are still plenty at the end of round 3. I note that, other than this one reversal between The_Jeh and 99of9 (by five whole rating points), the only difference between the new SoS and in-tournament WHR is breaking more ties. Looks very good so far. By the way, you are rating all in-tournament games equally, right? i.e. not entering dates for a changing rating and just reporting the final rating?
|
« Last Edit: Jan 26th, 2010, 4:20pm by Fritzlein » |
IP Logged |
|
|
|
Nombril
Forum Guru
Arimaa player #4509
Gender:
Posts: 292
|
|
Re: 2010 World Championship
« Reply #64 on: Jan 26th, 2010, 7:19pm » |
Quote Modify
|
on Jan 26th, 2010, 4:18pm, Fritzlein wrote:I note that, other than this one reversal between The_Jeh and 99of9 (by five whole rating points), the only difference between the new SoS and in-tournament WHR is breaking more ties. |
| ... the WHR puts me at 5th place while the SOS puts me at 7th place. Unless I'm missing something? Not that it makes *that* much of a difference to me - either way I'm looking for another upset!
|
|
IP Logged |
|
|
|
Adanac
Forum Guru
Arimaa player #892
Gender:
Posts: 635
|
|
Re: 2010 World Championship
« Reply #65 on: Jan 26th, 2010, 9:57pm » |
Quote Modify
|
on Jan 26th, 2010, 7:19pm, Nombril wrote: ... the WHR puts me at 5th place while the SOS puts me at 7th place. Unless I'm missing something? Not that it makes *that* much of a difference to me - either way I'm looking for another upset! |
| That was bizarre. My computer shows an SoS of 0.6945043921 for the three players ranked #5-7. And yet it somehow thought that yours was bigger than theirs Oh well, I found a solution and now the standings are in the right order.
|
|
IP Logged |
|
|
|
woh
Forum Guru
Arimaa player #2128
Gender:
Posts: 254
|
|
Re: 2010 World Championship
« Reply #66 on: Jan 26th, 2010, 9:59pm » |
Quote Modify
|
on Jan 26th, 2010, 4:18pm, Fritzlein wrote:By the way, you are rating all in-tournament games equally, right? i.e. not entering dates for a changing rating and just reporting the final rating? |
| That is correct, Fritzlein. The time factor is ignored. All games are asumed to occur at the same moment. Just one rating per player is calculated.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #67 on: Jan 27th, 2010, 6:22am » |
Quote Modify
|
on Jan 26th, 2010, 7:19pm, Nombril wrote: ... the WHR puts me at 5th place while the SOS puts me at 7th place. Unless I'm missing something? |
| SoS has you in a three-way tie for 5th-7th, so when the WHR places you 5th it isn't disagreeing, just breaking the tie. The only time WHR is disagreeing with SoS rather than just breaking a tie is raising 99of9 from 8th to 7th and lowering The_Jeh from a tie for 5th-7th to 8th.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #68 on: Feb 1st, 2010, 7:17am » |
Quote Modify
|
The upsets just keep coming. Adanac's win over chessandgo from 400 rating points down (and from a totally lost position) is comparable to Simon's win over 99of9 in the first round. It sure makes for a more unsettled, exciting feel than we had in last year's World Championship. I like it. Although woh and Hippo haven't played yet, I believe the top pairings for the final round are fixed at Adanac vs. Fritzlein and chessandgo vs. The_Jeh. 99of9 will play down against the top 2-2 player he hasn't played yet, but he has already played both Tuks and Simon, the top two. I believe 99of9's opponent will be woh if woh wins, and PMertens if Hippo wins. The rest of the pairings for the final round look complicated, and may require human intervention to sort out if the pairing algorithm fails to return a legal pairing. If I manage to beat Adanac in the last round of the preliminaries, we would have a three-way or four-way tie at 4-1, and the strength of schedule is currently close enough that any of the top three seeds for the preliminary could end up with the advantageous top seed for the finals, depending on how all the other games of the round go. Of course, if Adanac beats me he gets the top seed into the finals outright, while I drop to the #4 or #5 seed. I will wait for the last-round pairings to be finalized before I even go into the scenarios for the top 2-3 player qualifying or the bottom 3-2 player not qualifying.
|
|
IP Logged |
|
|
|
aaaa
Forum Guru
Arimaa player #958
Posts: 768
|
|
Re: 2010 World Championship
« Reply #69 on: Feb 1st, 2010, 9:33am » |
Quote Modify
|
The use of floating point scores is problematic insofar as genuine ties can be missed and no further tiebreakers will kick in.
|
|
IP Logged |
|
|
|
99of9
Forum Guru
Gnobby's creator (player #314)
Gender:
Posts: 1413
|
|
Re: 2010 World Championship
« Reply #70 on: Feb 1st, 2010, 12:47pm » |
Quote Modify
|
on Feb 1st, 2010, 7:17am, Fritzlein wrote: I believe 99of9's opponent will be woh if woh wins, and PMertens if Hippo wins. |
| woh and I have already played.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #71 on: Feb 1st, 2010, 4:04pm » |
Quote Modify
|
on Feb 1st, 2010, 12:47pm, 99of9 wrote: woh and I have already played. |
| Oops. I was trying to figure out why the algorithm paired you with Nevermind instead. Thanks for clearing that up!
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #72 on: Feb 1st, 2010, 4:20pm » |
Quote Modify
|
The pairings for Round 5 are up, and IMHO, we couldn't have asked for more intriguing matches. In the life-death matches we have 99of9 vs. Nevermind, PMertens vs. Tuks, and Simon vs. woh. All three games pit veterans against up-and coming players, so we can get a great read on how far the new wave has come. I can't wait for the games to begin! The games that determine whether we have seven, eight, or nine players at 3-2 or better are 99of9 vs. Nevermind and Hippo vs. Nombril. Specifically: nine if Nevermind and Nombril both win eight if one of Nevermind and Nombril win seven if neither Nevermind nor Nombril wins I'm afraid that Nevermind and Nombril each have to root for the other to lose, because if Nombril loses, Nevermind could also lose and perhaps still sneak into the finals as the best 2-3 player, whereas Nombril can win and still perhaps get left out of the finals as the lowest 3-2 player if Nevermind wins. By the same logic, all the other 2-2 players will be rooting for both Nevermind and Nombril to lose. Hippo and 99of9 have just earned themselves an automatic cheering section. Adanac will have to put his scenario generator to work to determine which players are most likely to be left out even if they finish 3-2, and which players are most likely to be included even if they finish 2-3. I can only guess that the way strength of schedule works will leave all of us hanging on the outcome of every game. Let the fun begin!
|
|
IP Logged |
|
|
|
Adanac
Forum Guru
Arimaa player #892
Gender:
Posts: 635
|
|
Re: 2010 World Championship
« Reply #73 on: Feb 1st, 2010, 4:51pm » |
Quote Modify
|
on Feb 1st, 2010, 4:20pm, Fritzlein wrote:Adanac will have to put his scenario generator to work to determine which players are most likely to be left out even if they finish 3-2, and which players are most likely to be included even if they finish 2-3. I can only guess that the way strength of schedule works will leave all of us hanging on the outcome of every game. Let the fun begin! |
| This table is similar to the ones from the past two years. The first 2 numbers show the best and worst possible rank after a win (taking into account all 256 permutations). The last 2 numbers show the best and worst possible rank after a loss. The most unusual lines are for PMertens, Nevermind or Woh: Nevermind, for example, can finish between 4-9 with a win but between 8-11 with a loss. It seems bizarre that he can advance with a loss or get eliminated with a win but both scenarios are extremely unlikely, at opposite ends of the SoS possibilities. In virtually all scenarios Nevermind will control his own destiny. PMertens and woh need a bit more help than Nevermind, but will advance with a win in the vast majority of scenarios. Fritzlein 1 3 3 5 chessandgo 1 3 3 7 Adanac 1 1 1 3 99of9 2 4 5 9 Tuks 4 6 8 10 ChrisB 12 14 14 16 omar 10 14 13 15 PMertens 5 9 8 13 The_Jeh 2 3 4 8 naveed 12 14 14 16 woh 5 9 8 12 Simon 4 8 8 11 Nevermind 4 9 8 11 Nombril 7 9 9 14 fritzlforpresident 14 16 16 16 Hippo 10 13 13 14
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: 2010 World Championship
« Reply #74 on: Feb 1st, 2010, 5:58pm » |
Quote Modify
|
Thanks for generating this, Adanac. It is truly bizarre that each of three players has an opportunity to qualify despite a 2-3 record or be eliminated despite a 3-2 record. What a crazy tournament! I note that even 99of9 can still be eliminated if he loses, so only the present top four are safely in to the finals. On the other side, though, the bottom five are out, as Hippo can't end up as the top 2-3 player even with wins by himself and 99of9. I take it that since Nevermind, PMeterns, and woh all control their own destiny except for corner cases, the brunt of not being in control falls mostly on Nombril. Even if Nombril wins, he needs either Nevermind to lose, or needs some help in strength of schedule. But despite being the most out of control, Nombril should still be rather better than 50% to make the finals with a win. I would be much obliged, Adanac, if you would re-run the scenario generator again as the early games of the round finish. It makes for great drama!
|
|
IP Logged |
|
|
|
|