Given the the relative lack of disparity in ELO rankings, particularly between 3 and 6 (and US is not far off at 2). I know it is hard to predict, but does anyone have a concern that this relative "closeness" may portend a lot of draws this Hex as teams are relatively close in capability (as defined by ELO numbers)? Mexico -- should qualify hands down and I am removing them from the conversation, although they are good for one or two surprise draws during qualification US -- Given form it is reasonable to think that a draw against Mexico is probable and about 3-4 draws is possible against the other teams in the Hex CR/HON/PAN/JAM -- I do not know how to truly separate these teams. On a given game day, any one of these teams can under- or over-perform.....In theory, playing to their abilities I would expect a high number of draws for these teams. MY point/query is about whether these models of predictions are reflecting expectations of too many clean wins/losses and not the muddling of draws that may exist between CR/HON/JAM/PAN given their relative sameness in the ELO rankings?
The model predicts two teams with equal ELO (adjusted for home/away) will draw about 28% of the time. I understand this is based on large sampling, and it seems to work. It predicts a level of total draws for the hex that is consistent with previous iterations.
Are you asking something like, "Donovan is not playing for the US, so their true SPI for this game is 72.2 rather than 72.8."? I think ESPN could do it, but only they have the data. This except from Silver's initial release explanation seems to imply that they could: "Any set of ratings can be combined in this way. Theoretically, they could even be used to assess the impact of particular players. For example, if David Villa (OFF90 rating of plus-0.38) were injured and replaced by Juan Manuel Mata in the lineup (OFF90 rating of plus-0.12, not shown here), we would estimate that this would reduce Spain's scoring output by about 0.26 goals per game, the difference between the ratings for Villa and Mata." I just read that whole piece again. Adding in the whole club ratings and game ratings is pretty complex. I bet in the inital run, it added some value. If those league/club/player ratings aren't really scrutinized and kept up to date they are going to make things worse though, and that's my fear with them. Regardless, since Dr. Gamera trusts them more, I ran the qualifying odds (including defeating NZ for the 4th place team) based on them. SPI says: Mexico = 99.7% Costa Rica = 78.0% USA = 62.8% Honduras = 62.3% Panama = 48.4% Jamaica = 16.5% Don't like that at all.
It's interesting to see them value CR over the US on the player pedigree, as CR only has ... what 3-4 players from the Top 6 leagues (?) and only two of them are full time starters (Brian Ruiz with Fulham and Joel Campbell with Betis). Brian Oviedo rarely plays for Everton (though he is not a bad player at all) and Junior Diaz is a part/full time swingman with Mainz. The rest are from the European "minors", CR home league, MLS.
Well. I guess the fact that they are running these means (I really hope so) someone is paying attention and keeping up the system. This is good news. My (and Reccossu's) attempt to convert SPI and run the sims did work out pretty well. My sim above had everyone's odds of WCQ% within 2% of what the Mothership got. I'll look at this and see if I can get closer to their results before the next time I run the SPI sims. Couple other notes: For today's match, they have Honduras win% at 54.8% and USA at 19.8% and they assume a+ 0.8 goals home field advantage. In my SPI sim and formulas, it comes in at Hon 50.4%, USA 27.1%. So, it looks like ESPN has a stronger home-advantage assumption (and possibly assumes a few more draws) than the +100 Elo combined with the formulas I'm using (which is probably correct, as I showed the historical home field advantages from the previous Hexes...although Honduras is the one place that historically under-performs at home). I like this color regarding the USA's ranking: "Despite notable friendly wins at Italy and Mexico in 2012, the United States' qualifying odds and SPI ranking (38th) are lower than many will expect, given the team’s reputation as a giant of CONCACAF. The low numbers are largely because of subpar results over the previous two years in the most important matches: World Cup qualifying and Gold Cup. This includes losses at Jamaica and home versus Panama, plus several narrow wins when heavily favored: a 2-1 win at Antigua & Barbuda and 1-0 home wins against Jamaica, Guadeloupe and Panama." I think this makes sense. What I don't really get then though, is why SPI loves Costa Rica. They went 1-1-2 in the Gold Cup and lost to Mexico by 3 goals. In World Cup quals, they lost to Mexico by 2 at home, and tied El Salvador (at home!). The only real "impressive" thing they did was beat up Guyana 11-0 in goal difference. It looks to me SPI is too easily influenced by running up goals against a minnow.
There are only 2 players that started in the Caribbean cup that will start tonight against Mexico and one of them is getting his start due to injury to another player.
Step 0ne, run. Says our chances to finish Top 3 now 62.1%, down from 74.4%. And it's going to get worse with the Elo drop.
I'm glad those ELO's didn't play the game for Jamaica against Mexico or else we would have lost by 6. In this group of 6 you can throw out your ELO , as it won't determine who gets to Brasil. Watch the ride.
Wonder what the number might look like today. In some ways, it works to our benefit that the other four teams tied. Points will be very difficult to achieve in this Hex.
I am glad ELO doesn't play the games, but wathcing the odds is its own game. With the three results from last night the US's odds of direct qualification dropped to 60%. Here's the current table (although not yet reflecting the ELO changes, just the matchday 1 results): Team-------------- Direct Q %-------------- Playoff %----------------- Out% Mexico: ---------------92----------------------- 5------------------------- 3 USA: ------------------60-----------------------19------------------------21 Panama:------------- 40------------------------21-----------------------39 Honduras: ------------46------------------------22----------------------32 Costa Rica: -----------48------------------------20-----------------------32 Jamaica: -------------13-------------------------13------------------------74
I think the ties work to the benefit of the US if it wants to finish first. But the ties hurt if they want to finish 3rd or better. The US might have been better off if Jamaica and Panama had lost, making it harder for those teams to end up ahead of the US.
If I've calculated the Elo points exchanged yesterday correctly, we have: Team, Top3%, 4th%: Mexico, 95.3%+3.2% USA, 55.7%+19.9% Honduras, 55.5%+20.4% Costa Rica, 44.4%+21.9% Panama, 37.8%+21.6% Jamaica, 11.3%+12.9% SPI is going to look freaking scary, as I'd think we'll be below Honduras and barely above Panama. I'm not sure they updated it yet. It says they have, and some teams have moved a bit, but Jamaica has the exact same rating as 2 days ago, which would not make sense to me.
Don't forget that the ratings ESPN displays are rounded off to the tenths digit, which can conceal rather large shifts. Suppose Jamaica had a defense rating of 1.04 going into the Mexico match, and a defense rating of 0.96 after the Mexico match; those would both display as 1.0. (Jamaica's attack rating presumably went marginally down, rather than up, as they were shut out.)
Problem is, we don't know if CR or Panama are the biggest threat to finish top 3. Your claim may be true if there's one team of those two who get a very low amount of points, but I think no team will finish with fewer than 9 points. In hex general, it may not be best to ALWAYS root for ties, since there are likely one or two minnows in the hex, and you'd prefer those minnow to score some upsets (wins). In this case, however, I think ties are the best way to go. Since there is no true minnow, we want points lost among all other teams. Draws do that best, IMO. As far as Jamaica, yes, would have been better had they lost. But that's assuming Mexico is a lock to qualify (grant, I still think they are the favorite to qualify anyway).
I checked the numbers assuming different results from the other games and it makes very little (less than 1%) difference in the US odds. Surprising to me, though, is that it looks like the tie in Panama is the slightly better case for the US. So, something good from matchday one.
I think getting one point as opposed to zero from Mexico in and of itself makes little difference. But I think the way Jamaica played and what it says about the quality and competitiveness of their team is a big deal. I'll be surprised if they aren't in the thick of things to the very end. This is going to go down to the last match day for every team except for Mexico.
When was the last SPI rated the Jamaican players? The Boys from Reading were certainly superior to their predecessors at their positions. If they can actually sit Ryan Johnson for Luton Shelton, they'd be a very dangerous home side for this level, having already proven that they could get a point on the road as well.
I'm not saying you're right or wrong, but in my prediction contest only 3 out of 19 people predicted everybody to have at least 9 points. The average predicted points for sixth place was 6.9.
Alright, I guess SPI is what it is at this point. You might want to close you eyes: Top3% + 4th%*WinNZ% Mexico = 97.7% + 1.7%*93.0%= 99.3% Costa Rica = 62.7% + 18.6%*74.8%= 76.6% Honduras = 61.1% + 19.6%*70.3% = 74.8% United States = 35.0% + 22.9%*69.2% = 50.8% Panama = 30.2% + 22.3%*65.9% = 44.9% Jamaica = 13.3% + 14.9%*52.2% = 21.1% It's a coin flip. Ouch.
My effort at converting the current SPI ratings and accouting for the matchday one results is about the same for the US, but really likes Costa Rica's draw in Panama: Team-------------- Direct Q %-------------- Playoff %----------------- Out% Mexico: ---------------88----------------------- 8------------------------- 4 USA: ------------------37-----------------------23------------------------40 Panama:------------- 20------------------------19------------------------61 Honduras: ------------57------------------------21------------------------22 Costa Rica: -----------79------------------------12-------------------------9 Jamaica: -------------19-------------------------18------------------------63 Bottom line is that JK better figure out how to put a coherent team together, because the sum of its parts is not enough.