It is, because you always have to know or judge or ask which data are important. A blind look at numbers without even knowing what these numbers mean is useless.
I never said look blindly. Everything that goes into a model should be challenged and its influence understood, but I do not think you need an expert level of understanding to succeed. It may help in some instances, or contrarily it could interfere. I don't know if you've looked at the link I posted above explaining SPI. It's primarily the same as Elo, with an attempt to use individual player data (derived from club data) to evaluate roster strength of past encounters to enhance the accuracy. Whether this addition is an improvement, or an introduction of noise is certainly debatable. One key difference with SPI is that it was built strictly in a predictive sense, while Elo is simply intended as a ranking system (though it certainly can be used for prediction as I have done in this thread). In this sense it "should" be a better system (though I'm certainly skeptical that it is). Ignore, it if you wish however. This quote from Silver certainly gives you the prerogative to do so: "Soccer is a rich, wonderful and unpredictable sport, and it would be quite a shame if a single number could tell us everything that we needed to know about a soccer team. SPI does not. It merely reflects the relatively limited statistical information that is available in international soccer, and does so in a way that is as fair and accurate as possible. In other words, SPI is designed to serve as a general guideline -- as a starting point for debates about team quality. It is not intended to be the ending point or to settle all arguments."
I didn't look at it yet, maybe when I have to waste some time more efficiently than here. However, with an academic background in computational physics, I know and like my numbers, but I also know that any numbers from any simulations can't be better than the basic analytical knowledge of the people who did the modeling; and applying successful methods from other sports or whatever to soccer is quite dangerous. However, I have to read more about it before I feel able to judge it. Anyway, I don't need numbers produced by a non-soccer fan as a starting point to talk about soccer. In the end I rather believe my eyes than any simulation anyway.
I'd say that for many people their eyes can mislead them because they hope for a certain result or fact to be true and, perhaps entirely unconciously, have stronger recall for data that supports that hope (if they are optomistic) or counters it (if they are pessimistic). I am not speaking of you, just generally. So Nate or other strictly objective assessments offer a control for that. Though, I agree an unbiased subjective assessment is probably more accuarate. I just don't know if I am capable of it. Regarding your UEFA qualifying example, I think we have a group with Germany, Denmak, Turkey, Ukraine, Serbia and Scotland. Germany feels good about finishing top three, but how confident is Denmark?
I don't care. Everybody knows that Germany will finish first, and everybody else competes for second, probably at least 6 points behind Germany. Right, from a pot-2-point of view, this is the group of death. It just depends on how many teams compete seriously for how many spots. There are more interesting things about that, like "how does it change with the number of teams and spots per group? How do the numbers change if you rearrange the groups?" and so on. However, I guess I'd have to read the linked article before I post again.
A very apt analogy there (although Germany should be Italy). Or how about SPI's version with this group: Netherlands, Denmark, Ukraine, Czech Republic, Slovenia, Belarus. Ukraine is fairly confident it can backdoor it's way in against New Zealand, but it's a battle to get there.
Using the last four Hexes, I went back to see how each team performed relative to how Elo projected they would, both at home and on the road. You can see the overall accuracy, as well as the home/road discrepancies below. United States (4 Hexes): Home Win Expectancy = 0.756, Actual = 0.875 Road Win Expectancy = 0.546, Actual = 0.475 Overall Win Expectancy = 0.651, Actual= 0.675 I shared this earlier, but for all WCQ, and it tells the same story. The overall accuracy is exceptionally good, but we outperform at home, and struggle away. It is somewhat the same story for the other two nations that have appeared in all four Hexes... Mexico (4 Hexes): Home Expected =0.810, Actual = 0.900 Road Expected = 0.612, Actual = 0.450 Overall, 0.711 to 0.675 That's a significant under performance on the road (although a couple of these may have been meaningless final games). Costa Rica (4 Hexes): E(H) = 0.635, A(H) = 0.800 E(R) = 0.413, A(R) = 0.375 Overall, Expected= 0.524, Actual = 0.588 A nice home field advantage for the Ticos. For the other three participants in this Hex, one has performed similarly to the above squads, one is a small-sample disaster, and the other is bass-ackwards. Jamaica (2 Hexes) likes the Home Office, while the road has been a bit tougher than they'd hope: E(H) = 0.506, A(H) = 0.600 E(R) = 0.268, A(R) = 0.250 Overall, Expected= 0.387, Actual = 0.425 Panama (1 Hex) had a rough first go: E(H) = 0.371, A(H) = 0.200 E(R) = 0.207, A(R) = 0.000 Overall, Expected= 0.289, Actual = 0.100 Honduras (2 Hexes) is odd, and has bucked the CONCACAF trend, both at home and away, but their overall performance is almost exactly what Elo would have thought: E(H) = 0.695, A(H) = 0.600 E(R) = 0.399, A(R) = 0.450 Overall, Expected= 0.547, Actual = 0.525 For Completeness here are the rest: T&T (3 Hexes): E(H) = 0.424, A(H) = 0.433 E(R) = 0.214, A(R) = 0.167 Overall, Expected= 0.319, Actual = 0.300 El Salvador (2 Hexes): E(H) = 0.355, A(H) = 0.600 E(R) = 0.174, A(R) = 0.100 Overall, Expected= 0.265, Actual = 0.350 Guatemala (1 Hex): E(H) = 0.429, A(H) = 0.700 E(R) = 0.283, A(R) = 0.100 Overall, Expected= 0.356, Actual = 0.400 Canada (1 Hex): E(H) = 0.569, A(H) = 0.500 E(R) = 0.340, A(R) = 0.000 Overall, Expected= 0.454, Actual = 0.250 Overall, in the last 4 Hexes, the Home Field Advantage has been quite a bit more than the 100 points given in the Elo formula. This shouldn't surprise anybody. Summing all the games, the home team performed at a 0.692 compared to an expected 0.607. El Salvador, Guatemala, Costa Rica, Jamaica, USA, and Mexico have all been tough to beat at home. T&T is around expected, while Panama and Canada simply were terrible in their lone appearances. Honduras is the weird bird, able to take unexpected points on the road, while dropping them at home when they shouldn't. I went ahead and used this data to create a home Elo and road Elo rating for each team in this Hex. I probably shouldn't because it's definitely too small a sample to rely on, and really over weights based on the performance of the past teams, but what the hell. Panama in particular gets hammered here because of their poor effort in 2005 (and I'm positive they will do better this round), but the others get adjusted more reasonably, and if their home/road tendencies from the past hold, here is what you get (with the unadjusted from earlier in ()): Top 3 %: Mexico = 97.6% (98.5%) USA = 76.4% (72.5%) Costa Rica = 44.1% (30.6%) Honduras = 35.9% (37.6%) Jamaica = 29.0% (21.7%) Panama = 17.0% (39.1%) That's kind of crap, a bit like a historical ranking really. Nevertheless, the home field, perhaps especially in Costa Rica's case (Saprissa or not???), should be considered a bigger factor when thinking about who will progress.
Good stuff. Maybe the 100 points ELO gives the home team is too little specifically for CONCACAF? I'd like to re-run mine with a "CONCACAF home" adjustment and see if it makes much difference. Accoriding to the formula posted earlier, the ELO difference between an expected .607 and .692 win percentage is 75 ELO points, so I will add that to everyone's home advantage and see what happens.
Not surprisingly, the change in home field advantage pretty much evens out. It doesn't even out precisely, though. That could be a flaw in my equations, or something real, but either way, it doesn't matter much. Direct Q% with additional homefield "average" advantage for CONCACAF: MEX -- 96% USA -- 68% PAN -- 40% HON -- 38% CR -- 32% JAM -- 25%
i'll take those odds. add in 4th place chances andit looks good i have hon over pan with pan beating new zealand i think any hex team beats new zealand despite how some kiwi fans feel
Yeah, was going to do this, but the home/home thing just cancels each other (it shouldn't be identical, so not an error). I calculated 181 for the Home Field advantage, BTW. So, I was going to post in the other "predict the Hex" thread. I ran a sim and then looked to see if their was any particular points finish that came up a large # of times, and I was going to use that as "my prediction". I think this is 1)Pretty cool math, but 2) an indication of just how even and difficult to predict this Hex looks to be. In the 100,000 runs, 94,940 were unique; that particular point combination among the six teams only came up once. There were 2,378 pairs, and 96 scenarios which occurred three times. The "large occurrence" was four, which happened with only four different point combos. This goes to show, predicting this hex is one tough cookie. I'm sure the scenarios would change each time (I ran one other and again 4 was the most, this time 5 different combos, all different from the first sim), but as these 4 made me smile/laugh a bit, here they were: Mexico 23, USA 17, Panama 14, Jamaica 12, Costa Rica 10, Honduras 8. Mexico 20, USA 20, Honduras 15, Panama 11, Costa Rica 10, Jamaica 8. Mexico 20, USA 15, Panama 15, 3 way tie for 4th on 11 pts. and my personal favorite... Mexico 20, USA 20, 4 way for 3rd on 11 pts. Goal difference sends somebody to Brazil, somebody to New Zealand, and two home.
Good stuff. I would need to add a little code to figure the most common point occurances -- I just keep up with how many times a team finishes in what position. might get to that just to try the prediction game, though. I do think it is odd that Jamaica ends up clear fourth in one of these -- very counter intuitive.
I updated the sim to reflect recent changes to ELO, including games played yesterday. Results: Team-------------- Direct Q %-------------- Playoff %----------------- Out% Mexico: ---------------96.5----------------------- 2.5----------------------- 1 USA: ------------------70------------------------15------------------------ 15 Panama:------------- 38-------------------------21.5----------------------4o.5 Honduras: ------------38------------------------21------------------------41 Costa Rica: -----------33-------------------------21------------------------46 Jamaica: -------------25-------------------------19------------------------56 Small differences only, but the US breaks the 70% barrier! Of course, it's not realistic that friendlies played yesterday actually change the odds of making the world cup, but they may help refine the ELO assessment of the strength of the teams and of what the percentages should be.
And using Nate Silver's most recent SPI, updated today: Results: Team-------------- Direct Q %-------------- Playoff %----------------- Out% Mexico: ---------------95------------------------ 3------------------------ 2 USA: ------------------51------------------------21------------------------28 Panama:--------------30------------------------21----------------------- 49 Honduras: ------------46------------------------22-----------------------32 Costa Rica: -----------63------------------------18------------------------19 Jamaica: -------------14-------------------------15------------------------71 Still not very comforting.
For some reason Nate has Costa Rica quite highly rated. I don't really know why. I don't see it, but his formula likes what their players are doing for their clubs I guess. Others may be able to explain it better.
Yep, small differences since the results yesterday were right in line with expectation. The best performance was ours with the draw away in Russia (+6 Elo). Honduras and Costa Rica got a nice but expected draws (+1, and +2), and Panama got spanked, unsurprisingly by Spain (-4). Reccossu, I'm not sure why are %'s are off a bit. I'm using the same formulas for W/L/D % and have built in checks to ensure the expected matches the sim. But I get a bit more favorable view for the US: Qualify for Brazil... Top 3 + 4th*W(NZ): Mexico = 98.4% +1.2%*90.1% = 99.4% (-0.1% from last sim) USA = 74.4% + 10.2%*74.8% = 84.6% (+1.4%) Honduras = 37.7% +22.2%*61.3% = 51.4% (+0.1%) Panama = 36.8% + 22.1%*61.4% = 50.4% (-2.4%) Costa Rica = 31.2% + 21.7%*59.0% = 44.0% (+0.6%) Jamaica = 21.4% + 19.2%*54.7% = 32.0% (-0.2%) New Zealand gained 4 pts by tying in China, so it's a bit tougher overall. Regarding SPI, it's a different system. Read the link I posted earlier in this thread and judge for yourself if you like the "enhancements". I will say that Nate Silver is most assuredly not involved with it anymore. This quote from Voros from over a year ago gives some insight into the system as well as what might be wrong with it: "Well, Nate isn't running it anymore and I actually very much like Nate's methodology and his using additional data like personnel can only help the results. Where I might criticize it is that the rankings are in their infancy and with things like this there are always needs to be a lot of tweaking as you adjust to how real life results diverge from your rankings. For example, I don't think the "home continent advantage" worked out very well in last year's World Cup, Ghana not withstanding." My fear is that no one at ESPN is paying any attention and things are out of whack.
Well, the simplest explanation would be that my program has a flaw in it. It almost certainly comes down to breaking ties correctly, and maybe my sim is biased in that unknowingly. Although the results are close enough that the flaw may not be obvious. When I get the energy I will try to deconstruct the program a little and see if I can figure out if I did something wrong. But I think we have a good idea of the basic range of expextations the teams should have.
how much money would nate silver put on usa not making brasil what is LAS VEGAS saying,, they're the guys putting bigbucks on ghe line i want a piece of the nate silver action, i'll put up $100,000 if he dos the same on usa making/not making it
Latest and last update before the Hex including recent ELO changes due to recent friendlies and the Centroamerica Cup. Results: Team-------------- Direct Q %-------------- Playoff %----------------- Out% Mexico: ---------------96.6----------------------- 2.4----------------------- 1 USA: ------------------71------------------------16------------------------ 14 Panama:------------- 49-------------------------22----------------------29 Honduras: ------------34------------------------24------------------------43 Costa Rica: -----------41-------------------------23------------------------36 Jamaica: -------------10-------------------------13------------------------78 Clearly ELO thinks Jamaica has fallen off the cliff. This is all due to poor results at the Carribbean CUP in December. I have no idea what kind of team JAM fielded, but Panama was the beneficiary in terms of ELO-based odds to progress. I will re-run the sim after the first round results.
Reccossu beat me to it. In my sims, Jamaica is punished even further. Coupling Jamiaca's fall with their improving Elos, Panama and Costa Rica are the big winners of this period. Honduras' head-to-head loss to Costa Rica in the UNCAF Cup final is weighted pretty heavily, which hurts the Catrachos odds. Jamaica's misfortune (French Guyana? Really? Martinique?) benefits the USA in that we finish in 4th more often with a weak-sister in the group. Here are my current probabilities for World Cup qualification according to a 100,000 sim: Top3% + 4th%*WinNZ% = Mexico = 98.5% + 1.2%*89.8% = 99.6% (+0.2% from last sim) USA = 74.4% + 15.0%*73.8% = 85.4% (+0.8%) Panama = 49.6% + 23.8%*64.6% = 65.0% (+14.6%) Costa Rica = 40.1% + 25.4%*61.3% = 55.7% (+11.7%) Honduras = 31.6% + 24.6%*58.0% = 45.9% (-5.5%) Jamaica = 5.8% + 10.1%*42.5% = 10.0% (-22.0%)
Which is why I lean more toward SPI than Elo -- SPI knows that neither country was fielding a first-choice team. Both SPI and Elo are miles better than FIFA rankings, though.
I love the idea of SPI, but see my comments a couple of posts up for my concerns about it. If you trust SPI more than you are very concerned, since our odds of qualification dip considerably (see the earlier posts on that as well).