Sep 22 10:00amEST —Survivor pathway 3B was modified.

Sep 21 7:30amEST —Early this morning, there were some erroneous fluctuations in some D/ST projections, while processing a model update. Done now.

Sep 19 2:30pmEST —Note that the Survivor pathway 1A has changed before TNF to include the Jets; The previous suggestion has now become 1B.

Sep. 17, 2024

News

• We are trying out a new “Basic” subscription tier option, for $7.50/year.

• Luckily my models did less-worse than others in Week 2— however the week brought chaos!

• The Details table for Kicker have a new breakdown, by field goal distances. Check it out!

Finally, I’ve written up this article about the background to Line Betting assumptions and strategy:

Betting Lines at Subvertadown -- Background, Assumptions, and Perspective

A few weeks ago, I started designing a video to try and better engage people in this topic. But I can see that will take too much time, and it's better to publish sooner whatever I can: which is this article! [After writing this, I see week 2 looks like a -50% loss, so perhaps well timed.]

My main goal here is to explain the level of expectations I have, from following the betting lines suggestions. There are assumptions that I make, and there's modeling to suggest what might be reasonable expectations.

.

Let's start with the basics.

Very tippity toppity first is: Only gamble if it's fun and entertaining. Subvertadown is certainly not a financial advisor, and as I've said before: if this were purely about making investments, then the advice would be: "Don't gamble!"

What we're doing. By assumption, you're betting approximately the same total amount each week, over the course of the season. (My models cover the 17 weeks in a season.) That means you're going to experience the highs of great weeks, the lows of terrible weeks, and mediocre weeks in between. It should be an advantage that spreading your luck over many weeks can minimize your risk of net loss.

How much? The assumed approach is that you set aside some amount of money, for the whole season. It should be an amount you are comfortable losing.

.

How much of it should you risk within a single week?

• Key background assumption: "It's gambling", and there's always a chance you could lose close to everything, each week. The probability of this happening affects how much you should save for later.

• Let's look at the worst case scenario. Assuming you wanted to bet $100 each week, the most conservative strategy would be to save that amount for each of the 17 weeks. That's the only way to be absolutely sure that there will always be enough saved to continue. Here's a graph of such a loss trend, starting with a $1700 pot:

Starting with 17x the weekly target is very highly inefficient, though, and that's because we don't actually expect the returns of each week to be 0. We usually expect to recover something, after all! So let's come back to this topic or weekly bet amounts, after we answer the next question.

.

What is a likely loss rate?

The Subvertadown models and methodology are based on plain bets that are assumed to have a 50%/50% chance of winning/losing. (You can of course find wagers where the expected probability of winning is, say, "70%", for a lower return on your wager.)

If every bet truly had a 50% chance-- and if those bets paid back a full 2x for a win-- then your average return each week would be exactly zero. That's because half your bets would lose and return zero, while the other half of bets paid back double. That would make the week-to-week graph of net earnings approximately flat-- meaning it would stay roughly constant, instead of dropping to zero like the situation in the graph above.

By consider this contrary situation-- where you essentially lose nothing-- it becomes clear that you very very probably don't need to save that huge $1700 for future weeks. If things were ideal like that, you wouldn't need to save anything at all for future weeks. You could simply start with $100 and end with $100.

However the returns from bookmakers is not a full 2x. The assumption made in the Subvertadown Betting Lines models is that the standard 1.9x is returned for a win (for a bet deemed 50-50). And the fact of not receiving 2x means that each and every week comes with an expected loss amount.

It's easy to calculate the expected average: Since you get 0 for the half of bets that lose, and since you get 1.9 for the half of bets that win, that makes an average return of 0.95 of the bet amount. And that means your baseline expectation should be to lose 5% of your bet amount, every week. The graph of expected loss, isn't completely flat, but it almost looks like that:

When you look at such a flat trend of returns, you'd probably conclude that reserving the full $1700 is a waste.

.

How much to save for later, if the loss rate were constant.

The reason that season losses appear very mild, above, is because of saving the full $1700, while only losing 5% of $100 each time. But you can clearly see from that graph that there is far too much saved, unnecessarily, for this idealized circumstance of constant loss. It appears that you'd be saving $1500 "too much" from the beginning! (In fact it's $1520 too much, but let's call it $1500.)

Subtracting that $1500 from the $1700 means starting with the much lower amount of $200. That would looks like this:

Note that the overall losses are the same as before-- it hasn't changed! (17 x -$5 = -$85) But things look more dramatic. To understand what's going on, you should consider the tradeoff between the (1) starting reserves and (2) the apparent loss %:

1. Starting with only $200 (instead of $1700) avoids "excessive saving" that you might prefer to avoid.

2. The losses "feel" much larger when you assume that you can re-invest bets. -$85 out of $200 "feels like" a 42.5% loss for the season!

This is meant to set your baseline expectations: with this level of re-investment, you should approach gambling with the idea that losing almost half (over the season) is the normal situation.

It helps to remember what has not changed though: the losses are still truly only 5% out of a total $1700 that got risked from multiple wagers, over the season. (It's adding up all the the green segments added up.) This is a reminder to scale your wagers to a comfortable level. For example, starting with only $100 could "feel like" it results in an 85% loss over the season. You might assess the need to recalibrate your approach, if that sounds scary.

How much to actually save for later

The current default suggestion from the Subvertadown "Save for later" is actually to assume that you start with this "$200". It's partly due to the considerations of the scenario I just described above. The underlying philosophy is that, if the weekly target bet amount is $100, I think we can probably avoid going negative (in some week) by reserving $200 (double of $100 ) at the season start.

However the story is more complex, because most season will always see a week in which a high loss occurs. Roughly speaking, there will typically be a week where a 75% loss occurs. The point I'm trying to make is that a 75% loss really isn't that much if you consider the scenario of starting with $1700. But as described, the 75% loss appears much more significant when you reduce the amount you save. So if a 75% loss sounds too scary, you might personally consider reserving even more from the beginning (thereby decreasing your weekly target amount since you reserve more for later). Again, you need to check with yourself what level you're comfortable with, meaning sticking to an amount that doesn't become stressful.

I would estimate that, if -5% was the true expectation of season losses, then a season start of $300 would probably be about right. The reason the Subvertadown approach does not recommend such a higher "Save for later" is because I expect most seasons will beat the -5%.

.

What are the expected results from Subvertadown suggestions?

Frustratingly, we really can't say, because the future is so hard to predict. The Subvertadown methodology definitely does not come with any guarantees.

But what I can point to is the combination of: opinion, historical results, and historical simulations worst cases.

• Opinion: I think it's more likely to break-even than to end up with a net loss, for the season.

• Historical results: The year 2023 did ultimately beat the break-even, ending with +$33 relative to the $200 start (you can call it +2%, following the theory presented above). Year 2022 ended with a loss-- I reviewed "the week 10 catastrophe" in my Accuracy Reports-- but I personally don't trust that year's result as indicative for the future, because the models were far too immature at that time. Still, I believe it helps to remember that a net loss can always happen.

• Historical simulations: It's easy to fall in love with numerical models, but it's absolutely essential to be skeptical that they are biased and don't produce as well in real life. So I'll instead just report the worst case outcomes, from simulating over the last 10 years. (1) 2014 would have been the year with lowest returns, but returns still would have been net positive, approximately +5% (so the opposite of the normal expected -5% loss). (2) The year 2015 also would have been a struggle, dropping to a low point of $125 in week 2-- later rallying in the final 5 weeks. This low-point of $125 supports the strategy of saving as recommended.

These points explain why my charts in my Accuracy Reports are labeled with "Low Case" and "Good Expectation" lines.

My goal of the above was to provide a simple explanation for what's going on with betting. If you're somewhat new to this, I hope you take away:

1. What are likely expectations for loss, and how to keep perspective over the full season,

2. How your "apparent returns" depend on the amount you reserve (or don't reserve), and

3. Why the guidelines are currently set up the way they are.

.

/Subvertadown