Week 10 Recap - What Happened

Welcome to the future, time to adapt or lose your bets.

Week 10 brought with it a much more palatable stretch of results that made me feel good about the human side of things here at OEF and even better about the machine side. And with that in mind, it feels like it is time to change the way this article operates.

Instead of focusing on the things that didn’t happen, it is time to transform this article into a series of autopsies that can help us assess our progress moving forward.

The thinking behind this is simple. In the first few weeks, we needed to examine more possibilities to see what trends we can identify. At this point in the season, however, including everything that didn’t happen along with everything that did creates a noisy and unproductive muddle of events.

Think of it this way: When you have only two points (i.e. two results), there is only one straight line between them, so we needed to create more data points to then see all the possibilities. Now, with so many data points, we need to remove some to make sure our line of best fit is predictive and not the results of non-real events.

In day-to-day life this theory may look like this: first time you do something, you obsess over all the potential outcomes. The millionth time you do something, you merely reference all of the previous experience.

Man, don’t you miss when this was just a football blog?

Well, no matter whether you understood the inner workings of a want to be stat nerd, know this: Monday recaps are now going to be (as mentioned above) an autopsy on the previous day with the hope that they can help predict moving forward.

This also signals a shift in the Wednesday and Saturday articles. Moving forward, we are going to turn the Wednesday article into a WORMS first look at each game, where we give you the WORMS scores and predictions heading into the week. The Saturday article will then be a human researched, WORMS powered, Best Bets.

To kick off this week of new articles, we’re going to take a look at what DID happen with WORMS, out missed losing bets, and our winners.

So, on that note, here is what happened this week:

What happened - WORMS dominated the spread

Of all the possible outcomes of WORMS, this is a 99% percentile job. In non-stat nerd speak, WORMS crushed it on Sunday.

Now a prerequisite here is that the Night Owl has already had to talk me down off a ledge when it comes to this success. I thought that we should bring this right to Bill Belichick’s door, but the Night Owl convinced me that we need more time to tell if this is blind luck or sustainable. I tried to retort that blind luck can buy you sight, but the Night Owl is more of a literal person.

Additionally, we are still learning how to apply WORMS to the betting markets, so keep an eye open for a tutorial coming in the WORMS section of the site as the evolves.

But for now, with tempered expectations moving forward, here’s how the first week of WORMS looked.

Out of 12 games (which includes Thursday Night but does not include Monday Night) WORMS went 10-1-1 against the spread and 6-6 straight up. There is reason to believe that the straight up numbers will improve, but the general pattern is not surprising.

First, an overall look at the slate reveals something pretty obvious: this was a wacky week. New Orleans, Kansas City, and Indianapolis all lost unpredictable games that very few saw coming. WORMS predicted that all these games would be much closer than their giant spreads indicated, but still sided with narrow wins for the favorites. In hindsight, these predictions appear to have been the most likely. Clearly these teams were closer than anticipated, but given 100 matchups we still might predict the favorites to win more often than not.

Which brings us to why WORMS was so good against the spread. When you get down to it, this is what the algorithm is designed to do. It is designed to process the most likely, average result and expose areas in which Vegas has overextended a line in one direction. A great example of this is the Miami Indy game. Looking at the numbers, Miami was not a team that should be a double-digit underdog. WORMS could not predict that Miami would win the game, but it certainly knew that Miami was being undervalued.

In fact, the only spread that WORMS lost was in a Detroit game that saw Jeff Driskel fill in for Matt Stafford. Outside of this real world variable, everything was right (with a push on Buffalo who we will get to in a second).

And so here is what happened with WORMS in its first week: it showed a propensity for being able to diagnose spreads, but less ability to pick winners. This reflects a logical betting narrative: it is easy to pick the most likely flow of a game, but not always as easy to predict which ways the balls will bounce. Picking close or not close is much easier than picking winners.

Of the winners that WORMS got right, four were in their top five for largest spread. Of the winners that WORMS got wrong, five of the six were in games that it predicted would be within three points. So when WORMS was confident, it was right; when WORMS was predicting a close game, it got the winners wrong. That makes sense.

WORMS did a great job removing as much chance as possible from the spreads, and we will keep tweaking and monitoring this going forwards. From my perspective though, this idea of correctly reading a game but losing a bet was very real this weekend.

And on that note...

What happened: I narrowly missed the Buffalo line and the Green Bay - Carolina over.

Part of the frustration with Best Bets this season is the lack of documentation about margin of error.

I have certainly been wrong by a lot (hello Green Bay and New England Week 9) but there is a large number of bets that are in the margins. Winning or losing a bet by a point here or there can feel as though you are getting lucky or unlucky.

For this week, the two bets that I missed were by a few bounces of the ball. I promise this article won’t turn into a justification of losses and a touting of wins, but it is valuable to explore for a moment the ebbs and flows of a bet to avoid getting too down on losses.

In the Buffalo game, things went exactly according to plan. There was a never a doubt that Cleveland would move the ball, but a strong of bad penalties, terrible play calling, and missed opportunities limited Cleveland’s scoring. On nine goal-to-go plays they scored zero points on one yard gained.

What ended up costing us the win, then was a Buffalo offense that didn’t produce and a kicker who missed two kicks. The most excruciating part of that formula was the Bills missing a kick as time expired. Having correctly identified this game as closer than the spread indicated, we then saw our fates rest on the 50-50 proposition of a 50-yard field goal. If that kick goes through we are in a better position to win the bet. As it stands, we lost based on our posted line and pushed based on the final line.

The other bet that was lost on the bounce of a ball was the Green Bay - Carolina over. Once again we were impacted by weather, but this was not what prevented us from losing this bet. That was actually the incredibly rare occurrence of two trips inside the five producing zero points (...the incredibly rare occurrence that also happened in Buffalo - Cleveland).

At the end of the first half, Green Bay elected to forego a field goal and attempt to score on the final play of the half from inside the Carolina five. They were stonewalled and the result was a deflated score.

Even with that, though, we were right there on the final play of the game when Christian McCaffrey drove to the goal line and was stopped/not stopped by an inch, depending on your perspective. The touchdown wasn’t awarded and what would have been a potentially game tying two-point conversion was not allowed.

In each game, the final play could have propelled us to victory, but instead was snatched from our grasp.

And that’s ok, as long as we get better at identifying these chance plays.

See, the goal of all of this is to avoid betting on these games that will end up being played on the edge of a knife. The Night Owl has suggested betting only on the heaviest favorites that WORMS produces. This week that would have been Baltimore, Green Bay, and Pittsburgh who all covered the spread and won outright.

And in line with that thinking, the two bets we won (pending Monday Night) reflected our ability to identify those massive gaps.

What happened: Miami and Baltimore crushed their spreads.

These are two very different teams and two very different stories, but with the same results: they beat a line that we identified as sketchy by more than 15 points.

In Miami’s case, Owl Eats Football came out as a pro-Dolphins publication and the results followed those beliefs. Miami was, for all intents and purposes, the better team on the field on Sunday. With Brian Hoyer as quarterback, people (and Vegas) were not viewing this as a close enough matchup.

And in this, we see what could be the most valuable lesson from this week - the nature of trend evolution over the course of a season.

During weeks 4-8 I made an absurd amount of money taking good teams against bad teams because Vegas couldn’t set the lines high enough. The Patriots were covering lines of 18 points with relative ease. Meanwhile, the Dolphins were consistently 15-point underdogs. What has happened, however, is that Vegas could not (or didn’t find it profitable to) adjust those lines based on a shift in that reality. Miami has now covered five in a row and won two in a row outright.

Spotting these shifts can be advantageous for us, and WORMS can help us spot the largest of these advantages.

Another area where the advantage was wide was Baltimore and Cincinnati. This was a situation where the line did not reflect the talent disparity and where we had this bet locked up at the end of the first quarter. While some teams zig and zag, the Bengals seem as though they are one a direct line to the bottom.

The point of touting these easy wins and decrying the narrow losses isn’t to look for sympathy or convince you that I am still worth reading, it is to highlight the shift we are trying to make in the implementation of WORMS.

Given that every game has an infinite number of outcomes, we are trying to identify the games that offer us the most winning outcomes. We want to stop betting on the single play differentiations and find those places where it would take a monsoon (hi San Francisco - Washington) to stop the outcome from happening.

After Week 10, we are much closer to honing in on bets like Miami and Baltimore and limiting bets like Buffalo and the Green Bay over.

If I lose my job to a robot in the process, so be it.

9 views1 comment

Recent Posts

See All