Prediction via Hollinger Power: Blazers to Finish 51-31 by: torridjoe Fri Jan 09, 2009 at 06:00:00 AM PST

A man with some interest in sports and numbers, and an Excel sheet, can be a dangerous thing. (Luckily I'm a trained professional.*) A popular parlor game among the Blazer faithful of late has been to look ahead at the rest of the season, and predict a level of wins that might be expected or realistically hoped for, as well as what the chances are that Portland makes the playoffs for the first time in what seems like forever. Most striking to me so far has been Gavin at 95.5 on the most recent Blazer podcast, going the other way from Casey and BEdge Dave on whether the Blazers would make the final 8 in the West. I think barring injury and some complete meltdown, the chances are very strong--like, 75%--to see Portland at least somewhere among the eight. But just guessing really isn't much fun, or at least my guesses aren't worth anything if they're just guesses. So I set about predicting wins the rest of the way based on some kind of rational analaysis. The short answer is that I look for the Blazers to win 29-30 more games of their final 47, to finish with at least 50...but not much more unless they really exceed expectations, either by total efficiency against the bottom dwellers or parity with the big boys. I used the Hollinger Power Rankings, a fairly stable and reasonably well-fed sample of data this far into the season. I noted the order and rating of each team, and also their record against Portland so far, as well as how many games the Blazers had remaining against them. Based on those two factors, I assigned an expected win percentage in four broad tiers among the 29 other teams, and then applied that to the Blazers' remaining games. The results displaced in my shoddy but endearing graphical way, below. {Step back into the 20th century below the fold!}

torridjoe :: Prediction via Hollinger Power: Blazers to Finish 51-31

Team Rating vs POR HRat Wins Pct% Rem. Adj Win     CLE 110.54 -6.72   0.3 2 0.6     LAL 108.39 -4.57 2.00 0.3 2 0.6     ORL 107.47 -3.65 0.50 0.3         BOS 107.2 -3.38 0.50 0.3         UTA 103.92 -0.1 2.00 0.3 2 0.6     DEN 103.84 -0.02 0.50 0.3 2 0.6     POR 103.82 0             ATL 103.56 0.26   0.5 2 1     NO 103.22 0.6 0.66 0.5 1 0.5     MIL 102.92 0.9   0.5 2 1     SAS 102.7 1.12 -1.00 0.5 3 1.5     HOU 102.42 1.4 -1.00 0.5 2 1     DAL 101.81 2.01 1.00 0.5 2 1     PHO 101.61 2.21 0.66 0.5 1 0.5     MIA 100.75 3.07 -2.00 0.6         DET 100.22 3.6 -2.00 0.6         TOR 99.67 4.15 -2.00 0.6         PHI 99.34 4.48   0.8 2 1.6     IND 98.01 5.81   0.8 2 1.6     MIN 97.91 5.91 -2.00 0.8 2 1.6     CHI 96.84 6.98 -1.00 0.8 1 0.8     CHA 96.59 7.23   0.8 2 1.6     NJN 96.16 7.66   0.8 2 1.6     NYN 95.76 8.06 -1.00 0.8 1 0.8     WAS 95.47 8.35 -1.00 0.8 1 0.8     GSW 95.37 8.45 1.00 0.8 2 1.6     MEM 93.94 9.88   0.8 4 3.2     LAC 92.33 11.49 1.00 0.8 3 2.4     SAC 91.44 12.38 -3.00 0.8         OKC 91.21 12.61   0.8 4 3.2               47 29.7                                                Curr 21 14              Fut 30 17              Exp 51 31       So, to explain. The first two columns are obviously the team and the Hollinger Power Rating. The next column is the difference between Portland and that team in Hollinger Rating. Minus means the team is higher than Portland in the rankings; positive means Portland has the edge. The "wins" category is sort of a hash; basically I counted the team's performance vs the Blazers so far, where they had played. If a team had won a single game so far, they got a plus one. If they played two and split, it was a .5. If they won two it was 2, if they lost 2 then minus 2. If they took two of three it was .66, lost two of three then .33. You get the idea. From there I looked at the numbers and drew up four tiers. The first tier is made of the six teams with the .30 noted in the win percentage column. That means that in any game against the Cavs, Lakers, Magic, Celtics, Jazz and Nuggets, I give the Blazers a 30% shot at winning any single game. The second tier are the even-up teams, where a split of remaining games seems likely, so each game is a 50% win chance. (Yes, I know that's not statistically precise since there may be cumulative odds at play, but it's an effective shorthand for this purpose). That group is made up by seven teams, and is--not unsurprisingly--mostly made up of teams battling Portland in the West in that 3rd-through-9th cluster. The third tier are teams that can be dangerous, but that Portland should be able to handle more often than not. I give them a 60% win chance there, although you see ironically the Blazers have played all six games in that group--and won all six. The performance against any one team or any small sample of teams is not so crucial; I don't necessarily expect that in any subset the number of games won will perfectly reflect the percentage I've assigned. However, in the aggregate I'm making the assumption that those ups and downs will even and cancel each other out, or be off only a couple games. The last tier are the bottom half of the league, essentially, and there is a noticeable dropoff from the top half. The first few teams--Indy, Philly and the Timberwolves--are relatively competitive (although the Blazers are 2-0 already vs MIN), but after the first couple of teams the ratings crater to 95 and below. Look how many matchups there are against the bottom feeders! I count 26 of 47 remaining versus these teams, including a whopping THIRTEEN games against the five worst teams in the league. Just under a third of the remaining schedule is Chump City! Of course they can't and won't win them all, but vs this group the Blazers are already 8-2, so 80% seems just about right. Give Portland 20-21 wins from this group alone, and then we're talking just 10 wins against the better half of the league, right around .500, to make 50. That's doable, isn't it? Look at the way it works out in the table and you'll see the concepts are really pretty conservative. There are four games against the Big 3 elite; the model gives Portland one of them--say, the Laker game at home. It also gives them just one win of the four left with Utah and Denver. Conservative, but definitely conceivable. Against their peers, the model gives them 6-7 wins in 13 tries against the likes of Atlanta, Dallas, Phoenix, the Spurs, Houston and Milwaukee. Some of those teams are mighty tough, but 6-7 is eminently thinkable. And then there are the 21 wins in 26 against the fatted calves... A couple words about how to read the table. The figures at the bottom are the current record, the predicted record the rest of the way based on the model, and the final record based on the past and the modeled future. Also, when looking at an individual game adjustment, discard any remainder. So for the Oklahoma City fourset, figure on three wins. In the aggregate however, all those bits of wins should be allowed to accumulate and become a sneaky win here and there, one where they beat the odds and maximize their record. Anything can happen the rest of the way, but to look at the very bottom-heavy tilt to the schedule, it's hard not to see at least 45, high 40s likely, 50 plus still a totally reasonable goal and--dare I say it--expectation. I also didn't factor in home vs away; by and large more games will be at home and I simply ignored that advantage, so the model hurts them for home games and helps them for away games. I don't think it has a major effect on anything. Again, look for it to mostly even out in the grand scheme of things. So, the Death Valley part of the schedule is over, and now it's time to come on back, Roy. There's no rest in the West; we need you! *no, really, I am.

Tags: win percentage, stats, pro basketball, ESPN, power ranking, Hollinger, Blazers, NBA basketball, (All Tags)

Print Friendly View Send As Email