Line movement, i.e. the difference between opening and closing betting odds, gets discussed a lot on our podcasts, and for good reason. The prevailing wisdom is closing lines are more accurate than opening lines in predicting future outcomes, and line movement is indicative of how much money is being wagered on teams and/or players. If a golfer’s odds move from 225-1 to 75-1 during a week, there’s enough belief out there, backed by actual wagers, that that particular golfer is being undervalued. All else being equal, since closing odds are more accurate than opening odds, we would therefore expect a golfer with good line movement to perform better than a comparable golfer with no line movement.
I’ve been on board with the idea of incorporating line movement into picking players, but I haven’t seen good rules on exactly how to incorporate it, mainly because they don’t exist. The details aren’t trivial, either. What if a low-rated player in your player model has really good line movement? Do you rate that player a little higher as a result, or does line movement trump your model and they become a must-play? And exactly how much line movement would a player have to have to make them a must-play? The flip side can get equally confusing. If a highly-rated player in your model has good line movement, should you take even more exposure on that player, or were you just properly valuing that player in the first place and the market is simply catching up to your valuation?
Since no one else has tried to answer that question, I decided to make the answers myself. Those answers start with quantifying just how much line movement can tell us about future performance. I collected opening and closing odds-to-win lines for every PGA tournament in 2016 so far, then linked that data to their actual results. There are a lot of ways to measure performance, but we’ll start with a super-simple measure of performance to start: How many birdies a golfer scored relative to the tournament average. This is a crude way of adjusting for the scoring difficulty for each course and gives us close to an apples-to-apples comparison of line movement effects of each course.
To evaluate line movement, I started with a very simple model: Predict a player’s birdie performance based on their opening odds to win. The higher the odds, the more birdies we’d generally expect them to score. Here’s what that simple model looks like:
So for each tournament, the difference between their actual birdies and their expected birdies is a measure of each golfer’s over/underperformance relative to expectations. In order to measure how much line movement matters, we’ll plot each player’s line movement against that difference in performance level. If a player has really good line movement (i.e. his closing odds are higher than his opening odds), we would expect him to do better on average than whatever we predicted based on opening odds alone, right? Here’s how that plays out in reality:
The correlation there is . . . almost zero. (In fact, if you remove Rory’s round at Wells Fargo, the trend becomes outright negative.) Not only was I not expecting this in the slightest, but it’s forced me into potentially my hottest take yet: Open-to-close line movement may not be all that predictive of DFS outcomes.
I don’t think this is conclusive by any stretch. Line movement has predictive power in literally every other sport, so I’m not exactly ready to throw away the concept for golf based on one simple analysis. Here are some explanations for why the data shakes out the way it does:
– Not all line movement is created equal. Lines move on multiple occasions during the week, each in response to different bets. Sometimes dumb money can cause lines to move just as much as sharp money, but line movement treats the two as equal. The trick is knowing which line movement comes from which group. This is 100 percent anecdotal, but line movement tends to be more sudden later in the week, and only large bets can move lines that quickly, which is more in character with betting syndicate wagers rather than public wagers. That graph above may look better if line movement were defined as the movement from Tuesday to Wednesday instead of Monday to Wednesday, but there’s no way to know for sure without capturing multiple snapshots of lines.
– Line movement at different opening odds may mean different things. A golfer moving from 200-1 to 100-1 has roughly the same line movement as a golfer moving from 100-1 to 67-1. And it’s the same for a golfer going from +1000 to +950. Should we give the same boost to all three golfers, or do your initial odds matter? It’s an answerable question, but the analysis gets a little more complicated and non-linear. And if it does matter, it definitely paints a more complex picture of what line movement is and when and where it matters.
– Line movement may actually matter less than we may think. I also don’t want to discard the hot-take answer quite yet, at least in a weaker form. Odds-to-win bets are just that, not “odds to do better than DK prices.” There’s still a hypothetical golfer out there who isn’t a threat to win necessarily, but is a pretty good cut-maker and represents solid DFS value. That golfer probably won’t pick up a lot of line movement, but is that a reason to not put them in your lineups? In the same vein, a boom-or-bust longshot golfer could see great movement for logical reasons but would be a terrible cash-game play. What if you had to choose between those two golfers to round out your cash-game lineup? The former is probably the better option, so you could say credibly that line movement doesn’t matter in this case.
The only hard conclusion I have at this point is line movement evaluation is a work in progress. I’ll be collecting more information over the course of this season to reflect the findings, and at the end of this series, I’ll showcase a methodology I have for evaluating these line movement-vs.-model value tradeoffs I talked about earlier to try and clarify the DFS decision-making process.
