Wednesday, 23 May 2012

Rating Statistics

Heat Big Three, Celtics Big Three
Scott sent me a link to a good piece on ESPN about the technology that some NBA teams are using to track their players. Well worth a read if you are interested in this topic, but for me, it does beg the question of how useful this information, and indeed other similar data from 'fluid' sports, really is from a gambling perspective.

Baseball, and its sabermetrics, is atypical of most sports, in that it is a game made up of set plays, all of which start with a pitch, and with a limited number of starting and ending positions. Thus over the many years that statistics have been tracked, the historical probability of team A winning from a position of being x runs ahead after y innings with runners on first and second with one out (for example) is known. More importantly, because the game is made of of set plays, there is enough time between plays to actually use the data for betting. Unfortunately, in-play baseball liquidity in most games is rather poor, but in theory you can gain an edge.

Spurs Big Three
In basketball, not so much. Yes, the game stops often enough for time-outs, or free-throws, but when a game is in full flow, I challenge anyone to be able to make use of rebounding statistics, distance a player has run during a game, top speed etc. Valuable information for the coaching staff I'm sure, but outside of being able to use free-throw statistics (it makes a big difference which player is on the line, and late in a game, can make a big difference to the probability of a team winning) the data is of academic interest only to punters. Jamal Crawford hits 92.7% of free-throws, Dwight Howard hits less than 50%. Three point shooting is another useful statistic - I've mentioned the Golden State Warriors and their ability to hit three pointers, and last season they had 3 players in the top 20 in this category (plus a fourth who is now at the San Antonio Spurs). In fact, the Spurs are even better, with five of the top 20 players in the league. The Charlotte Bobcats best three-point shooter, I hear you ask? D J Augustin - 34.1% and 94th place.

For football, there are some great blogs out there that analyse all kinds of aspects of the game - headed goals, woodwork strikes, etc. including, rather perfectly timed, this piece from the excellent - which mentions:
I'm not writing this post to demonstrate how awesome my little algorithm is, but rather to demonstrate that predicting league performance, whether by on-field event data or off-field financial data, is very difficult and captures very little of the variance present in actual league performances.
As a gambler, it's important to be able to separate the useful data from the interesting data. Is team data more important than individual data? For a sport such as basketball, clearly the individual player data is of greater importance than in an eleven player sport such as football. The loss of LeBron James for example will impact the odds on a Miami Heat win far more than the loss of even a missing Messi would be for Barcelona. Most top NBA teams have a so-called 'big three', and the loss of one or more of these can turn a 1.5 shot into a 3.0 shot, as we have seen this season with the San Antonio Spurs. One reason why rating NBA teams purely on team statistics is fraught with problems. As for baseball - the same teams play against each other on two, three or four consecutive days, and the prices fluctuate wildly, all because of one key player - the pitcher. To be meaningful, team ratings need to allow for the stats for this one individual. Unfortunately, starting pitchers rarely pitch the whole game, so how much does the starter's numbers count for? Perhaps not so much as you might think, and perhaps a reason why opposing short priced favourites in baseball is a better option than trying to buy money. 

Peter Nordsted has certainly had some success with this strategy this season, and I believe last season too (at least until the late part of the season when favourites, often 'needing' to win, seem to do just that a little more regularly), and he also offers one good reason for betting based on ratings, even if he does end his sentence with a preposition:
That's the beauty of following a value ratings based system. It takes all of the emotion out of making the decision of what to bet on.
I remember clearly my Mother overhearing me tell someone to "Fuck off" when I was a young boy. She was furious. "Don't you ever let me catch you using that expression again", she shouted. "Using language like that is not appropriate. 'Off, is the direction in which I would like you to fuck' is much better English" - a lesson I have not forgotten.


BigAl said...

I think you're probably underestimating the skills of the very best sports gamblers / modelers. This sort of data would be invaluable to them. Obviously not all of it, but they'd love the chance to find out.

In terms of in-running, you'e right, probably not that useful in the current markets. But mainly for liquidity reasons.

A guy called Haralabos Voulgaris is an interesting character in the NBA betting world.

BigAl said...

(In first paragraph last sentence i.e. to find out what is useful and what isn't - by having access to it)

Pete Nordsted said...


I have now started blogging again at and I have began by mentioning a post you put up in September 2009 that really changed my approach to betting especially on the US Sports.

I would be very grateful if you could put a link to the blog on your blog.

Keep up the good work