[vicwootn]

i've been pricing NBA a lot lately and i found i have a TON of unders. this happens for NFL too. i want to say at least 80% unders. should i find 50/50 overs and unders? maybe a few more unders because fish bet overs?

this got me thinking about using median instead of average. that way i'm not factoring in 2 point NBA games that Carmelo Anthony has. is the only reason you use average is to price props very quickly? a quantity vs quality argument? for NFL i'm able to look through passing yards very quickly to see if there is any low yardage games i should discount or adjust for. for NBA it's not quite as easy due to more games. is there any way you adjust when you price for either sport? the following is an example that was the nail in the coffin :


carmelo anthony

23.9 season mean +1.1
25 season median

26.3 10 games mean +3.2
29.5 10 games median

23.7 20 games mean +2.3
26 20 games median


is season mean really the best to use hands down when normalizing NBA player points? would you ever consider 10 games median or any other of these? what about weighting season mean and 10 game median together? 1.1 point difference between season mean and median is enormous! i know this is a small sample but here are two more examples :

dwayne wade

25.1 season mean +0.9
26 season median

29.4 10 game mean +1.6
31 10 game median

29.25 20 game mean +0.25
29.5 20 game median



lebron james

25.4 season mean -.4
25 season median


28.6 10 game mean -3.1
25.5 10 game median

26.6 20 game mean -1.1
25 20 game median

kevin love was a 2.3 point difference between mean and median.

lastly, is there a minimum edge you look for when calculating props to overcome these sample errors?

[Justin7]

A lot of oddsmakers take the mean, and set that as their price. This is frequently more than the median or the expected mid-Poisson result.

Consider a receiver who in 6 games, had this many receptions:
1
1
2
2
3
9

The median is 2. The mean is 3. Bad prop oddsmakers would set the over/under at 3, and you'd bomb away on the under every time.

re: NBA. Look for big changes in the numbers. If a guy's playing minutes jump up a lot in the last 10 games, there might be a change in the coach's lineup strategy. I often just use the 10 most recent games, and only if the number of minutes is fairly constant in each game. It's much harder to price if a player's minutes doubles at some point.

[vicwootn]

re: NBA. Look for big changes in the numbers. If a guy's playing minutes jump up a lot in the last 10 games, there might be a change in the coach's lineup strategy. I often just use the 10 most recent games, and only if the number of minutes is fairly constant in each game. It's much harder to price if a player's minutes doubles at some point.

so it would be wise for me to use a 10 game median (instead of season average) for normalizing provided the minutes are constant?

[Justin7]

I would look at both. You'll usually find that the median is a bit lower than the average. If that is the case, I would use the median as the "correct" number, especially if the average is 5-10% higher.

[vicwootn]

thank you!

obviously you would take the over if both the average and median were over. would you take an over if the median is over BUT the average is under or just not bet?

[Justin7]

thank you!

obviously you would take the over if both the average and median were over. would you take an over if the median is over BUT the average is under or just not bet?

As always, it depends... But yes, more often than not.

[vicwootn]

As always, it depends... But yes, more often than not.

can you provide an NBA points example where it would not be a good idea to bet with the median over and average under? thank you