Watching the TAB markets as they've shifted across the course of a week I've often wondered if there might be something predictive in those movements. If the eventual favourite's price has shortened during the week, does it win more or less often than its closing price would suggest?

Until now I've not had the data to investigate this question, my TAB data being reliably captured at only one point during the week, but the historical opening and closing Pinnacle Sports odds data, made available in a spreadsheet from here on the AusSportBetting site, has made such an analysis possible

(In the "Terms of Use" for the spreadsheet there are the usual caveats about the possibility of errors, which we should acknowledge and remember as we interpret the results of the analyses that follow, but a scan of the data suggests that there are no major, obvious deficiencies.)

THE ANALYSIS

For today's analysis I'll be using the opening and closing Head-to-Head market Home and Away columns from the spreadsheet. Mostly, I've taken the data as read, though I have excluded those games where the compiler of the data noted that he or she has used synthetic data because Pinnacle Sports chose not to field a head-to-head market for a game. That leaves me with a possible 592 games to analyse from across the period spanning Round 1 of 2013 to Round 22 of 2015.

For the first analysis, I've classified every game on the basis of whether the closing prices, relative to the opening prices, represent a shortening, easing, or no change for the favourite and underdog in the closing market. So, for example, if the favourite in the closing market opened at $1.11 and closed at $1.08, it will be classified as having shortened. If the underdog in that same game opened at $7.08 and closed at $9.84, it will be classified as having eased.

In total then a game could wind up with any of nine different classifications, though I'd expect few games to show a shortening in the prices of both teams since this would imply an increase in the Pinnacle market's overround as game time drew nearer, which is the opposite of the behaviour I've experienced with the TAB.  

The table below provides counts of the number of games in each category and reveals, for example, that in only 16 games did the prices of both the favourite and the underdog ease. The most commonly observed behaviour was that the favourite's price shortened and the underdog's price eased. This was the case in 326 of the 591 games, or just over 55%. (Note that I've ignored the one game where the closing prices had both teams at the same price.)

Next most common was an easing in the favourite's price and a shortening in the underdog's, which occurred in 225 games, or about 38% of the time.

As expected, few games - just 1, in fact - saw both team's prices shortening.

Also recorded in the table is the ROI from level-stake wagering on the favourite or the underdog in each game classification. For example, wagering on the 16 favourites in games where both the favourite and the underdog eased would have produced a +3.2% ROI, while wagering on the 16 underdogs in those same games would have produced a -49.0% ROI.

The only potentially meaningful ROIs in this table are those for the two most-common classifications are these show that:

• Wagering on the 225 underdogs in games where the favourite's price eased and the underdog's shortened earned a +3.7% ROI. Wagering on the favourites in these games had an ROI of -3.3%.
• Wagering on the 326 favourites in games where the favourite's price shortened and the underdog's eased earned a -0.1% ROI. Wagering on the underdogs in these games had an ROI of -4.9%.

In general then it seems more lucrative to bet with the weight of money - that is, on the team whose price has shortened - but specifically when it's the underdog whose price has done the shortening.

Wagering on underdogs almost inevitably produces a volatile stream of returns, and that is indeed the case if we wager just on underdogs whose price has shortened when the favourite's has eased.

Given such volatility, some caution is required in interpreting the apparent profitability of the strategy.

SUMMARY AND CONCLUSION

There does appear to be some evidence, in the Pinnacle market at least, that changes in market prices provide a useful signal about the probabilities of the participating teams. Based on the small sample size we have here, wagering on underdogs whose price has shortened when the corresponding favourite's price has eased appears to offer the potential for profitable wagering.

Due to the volatility of the returns, I'd want to see a bit more confirmatory data before I acted on this, however, but it does appear to be a phenomenon worth tracking and, if possible, exploring with another data set.