W’s and L’s

日期:2014/07/04

"The people who are coming into the game, the creativity, the intelligence—it's unparalleled right now. In ten years if I applied for a job, I wouldn't even get an interview" -Billy Beane quoted in The Signal and the Noise by Nate Silver. Silver knows baseball very well, and there are many insights and carry overs in his chapter on W's and L's in his book. Here are some of them.

1. Silver developed a system called Pecota to predict when a hitter was going to be good. He picked Petroia who became a Most Valuable Player whereas all the other systems missed him. He started out badly and then improved greatly. The principle of ever changing cycles applies to baseball as well as our field. While silver doesn't know anything about markets and his chapter on it is one of the worst I've read, he seems like an amiable personage. I like his humility. He goes up to Petroia to get an interview: "'No. I won't give you a minute. I'm trying to get ready for a Major League baseball game,' he said in as condescending manner as possible every syllable spaced out for emphasis."



 2. In developing his system, he tries to weasel out skill from luck the same way Galton did. He doesn't like batting average but likes things like home runs and strikeouts versus walks. Would we be better by looking at how far down or up, the market was rather than the win or loss.

3. There is an aging curve. A player is good after a few years but bad near the end. It's sort of like the s curve for growth. Silver tries to capture which part of the aging curve a player is on and uses that to pick how much to pay a player. He doesn't seem to realize all the difficulties in differentiating between the 20 kinds of curves that are possible, and the predictivity of making assessments even if you knew the curve a player was on. It's very similar to the problem we have in looking at similarities. Which are the variables to measure, and even if you could find the most similar would that be predictive. Neural networks is based on the similarity algorithms.

4. Bill James comes up with a similarity rating starting with 100 to see how a player compares to other greats. Seems to use linear distance. Much of James work should be applied to markets. The trend follower who lost so much who's now the baseball owner should have used James as his chief speculator rather than following blindly the moving averages.

5. Silver concludes that scouting + computers is better than just computers. I believe that no system is good without judgment and the question of clinical versus objective rating is a ongoing debate in psychology.

6. Silver actually evaluates his predictions versus the scouts and concludes the scouts did better. One has to compliment him for his objectivity in making such an evaluation. It is amazing how few of the forecasters in our field actually provide an evaluation.

7. Silver points out that everything about baseball is encapped in the score cards and the videos. He believes that baseball is the best slate, the most detailed and accurate base of operations for forecasting. I wold say that our own field where tick data for all trades is available is just as good.

8. He comes up with 4 factors that go beyond the statistics that are good for evaluating a player. Preparedness, concentration, competitiveness, and stress management. It would be good to have Brett's take on these factors. They all seem reasonable and might be applicable to our field in choosing employees and partners but they are untested. I would think humility would be one of them for our field as well as hard work. I like that Petroia never wasted a second but tries to play his hardest even in the warmup. That's what I like in a trader and how I tried to be in racket sports also.