Tuesday, May 17, 2011

Data on PickSix

A few weeks back, Ottoneu's new daily game was released, and it has quickly grown to over 200 users. Niv Shah sent me some data on the first 12 days of the game (points by position for every entry for each day, totaling 1,632 entries) and I took some time to breakdown some of the numbers.

For starters, I wanted to see if Niv succeeded in creating what he hoped to create - a high variance, daily game that would allow anyone to get involved by having a wide range of possible outcomes. Well, over those 12 days, the average score was 30.8 and the standard deviation was 17. This means, assuming a normal distribution of scores (which appears to be roughly true) that 95% of scores will be between -4 and 64 - but that there is definitely room for scores ranging higher or lower.

In fact, the lowest score seen in the first 12 days was -18.1 (largely driven by Huston Street imploding for -18.7 points, but still, .6 without your reliever is nothing to be proud of).

The highest was 108.7. Actually, in the first 12 days, only two people scored over 100 and both were on the first day. Both rode a combined 33.5 from Zobrist to that total, although one of them got 39.1 out of a single player as well - Lance Berkman.

Among individual players, the averages and ranges differ slightly by position.

Catchers averaged 4.8 points with a standard deviation of 7.3, and ranged from a low of -5 (basically, an 0-X from a position player gets you a score of -X) to a high of 31(courtesy of Victor Martinez on May 9.

Corner Infielders were the best position players on average (5.3) and also had a standard deviation of 7.3. the low here was again -5 and the high was the previously mentioned 39.1 from Berkman. By the way, that 39.1 on day one of Pick Six was the highest score by a single player over those first 12 days.

Middle Infielders were the worst position players (not a surprise) with an average of just 4.1, but again had a standard deviation of 7.3 - the variance has been rather steady. The low score for a MI was -6 and Zobrists 33.5 was the highest.

Outfielders were paced by the same 39.1 from Berkman as Corner Infielders, and were second best overall at 5.2. They also had slightly more variance (7.6) but the difference is basically meaningless. The low score was -6 as well.

Pitchers did not have quite as high variance - 3.9 for SP and 5.3 for RP, but something interesting is going on with RP, which I'll explore in a moment. SP averaged 7.9 points per entry - easily the most valuable position, but their low standard deviation suggests a limited upside - which proved to be true. The highest SP score was 16.5, a May 4 effort by C.J. Wilson. The lowest was -6.6 out of Ryan Dempster on April 28 - a start used by just one Pick Six user.

As I mentioned, there is something odd going on with RP. The average for relievers is just 3.4, lower than any other position, but this is LARGELY driven by RP not playing. Every other position you can find out before the game if the guy is playing or not - RP are a bit of a leap of faith. When you eliminate the 876 relievers who didn't enter games, the average jumps up to 7.4 - nearly in line with SP and above the position players. The standard deviation barely budges to 5.5.

A few other interesting notes:
Already covered the highest scores overall, but the worst position player scores were Dustin Pedroia on 5/4 and Brennan Boesch on 4/30, both with -6. I feel for the users who picked Pedroia, but Boesch? Sometimes you get what you deserve.

The correlations between total points and points from any given position range from .47 to .50 for position players, but are just .22 for SP and .34 for RP. I haven't explored this fully yet, but I believe this is a result of users picking position players who are facing bad pitchers - take 2-3 Indians when they face Kyle Davies as happened last night, and you are going to have a lot of success. I think a lot of people use that strategy (I will have to look that up somehow) and so when that pitcher really does struggle, you end up not only with a good score from your OF or MI or CI, but from all of them - leading to a high total score. Of course this also works the other way - that pitcher steps up, you are in trouble.

Interestingly, I have avoided this strategy, feeling that I didn't want to risk having a really bad day, just to take a shot at a great day - besides, no reason I can't have a great day picking players from multiple teams.

Well, it seems I was right. My average score since day one is 37.6 and my standard deviation is a bit lower than the overall, at 14.9, so I have had slightly less volatility than the average user. As I have solidified that strategy, the difference has become more stark - my standard deviation the last 10 days is just over 10 (note that I do not have overall data for the last ten days, so it may be that EVERYONE is getting less volatile).

So what does this all suggest? Well, if your goal is to play for the long-haul - have a high average and brag about your consistency - you probably want to diversify your lineup and go with safe RP. You can accept a missed night from your RP but not a Street-like collapse.

But if you are playing for glory and greatness - if your goal is to win a day and have everyone look up at you on the leader board for those 24 hours - make sure you find a RP who is likely to pitch that night (hasn't gone the past couple days) and take a risk on an offense going up against a pitcher likely to stink up the joint - it is probably your best bet for a big night.

1 comment:

  1. What happened to your blog Chad? No posts in the last six weeks!