There's a fairly clear trend then, in that games with 3 or fewer turnovers go under the line more often than not, whereas games with 4+ turnovers go over the line. The correlation is quite strong, so it seems reasonable to conclude that turnovers typically lead to a higher scoring game. There were a very small number of match-ups that produced no turnovers from either side, and those games went under almost 60% of the time.

Good enough then, the second test we could run would be to see how the season-to-date turnover numbers coming into the game for both teams in a match-up compared to the eventual over/under outcome. To do this we add together the turnovers committed (per game) for both teams:

So there's a very modest predictiveness to the above. Basically teams which have not committed many turnovers coming into a game are slightly more likely to play over the number, whereas teams which have committed many turnovers are slightly more likely to go under.

We also looked at it using "turnovers forced" and both "forced and committed" but to be honest the results were no more interesting than the above. The final test we want to run then is to look at the difference in turnover history between the two teams:

Again, this would unfortunately appear to be "noise" -- not a lot we can take away from this. The last step we're prepared to take, and even this risks the dreaded backfitting alarm, is to examine these T/O difference cases with regards to specific line ranges -- we'll use three line breakouts: 37 or less, 37.5 to 42.5, and 43+

Well here at last we start to see something that might be helpful -- when you find games with an extreme turnover difference number of 1.5 or higher (which means that one team say is -1.5 turnovers per game and the other is +1.0 per game), then there have been some trends in the low line and high line totals. In general of course playing over on low lines and under on high lines has been the historical smart move, so that's a part of the reason this breakout looks good.