*The Potential of the Casino

[iStegosauruz]

[div align=\\\"center\\\"]Methodology[/div]
Let’s talk sports betting. In the last few days I’ve learned more about spreads, lines, prop bets, and odds making than I ever expected to. When the casino was announced I was excited – I think the league needs more ways for players to spend money, especially considering how available it as a resource. Conceptually, a casino provides us as players to use our money but also provides a way for the league to siphon money out of the economy, increasing the scarcity of the resource and making it a more valuable commodity. That all being said, when I saw the current options to bet on in the casino, I was left feeling fairly underwhelmed.

For the casino to function as it is intended to, the bets need to be enticing enough to allow players to feel excited about the options while also balancing the need to have some players win, some players lose, and money to be siphoned out of the economy as well. If the system is set up properly there should exist a sweet spot for those goals. I set out to find out the proper way to set up spreads and over/unders with compelling odds.

I’ve created a model that uses 32 different variables, 32 coefficients, and 1 constant to produce a starting value for point spreads and over/unders. The model utilizes the data from last season to calculate the variable values needed but I’m working on incorporating previous seasons. I will also be including the data from this season as games get played. What this means is that the model essentially “learns” and improves its projections the more data it gets. These projections are just starting points – whoever is running the casino can adjust as needed for various factors.

To go along with that I’ve created a standalone algorithm to produce odds for the spreads and over/unders the model formulates. The algorithm factors in five different pieces of data to create the projected odds. It is key that the model and algorithm remain separate so that neither influences the other. Once the odds are produced by the algorithm, they can be smoothed over to make them more digestible numbers and then filtered by which ones should be used on a given week. What I mean by this is that not every game produces odds that are worth betting on, however with six games a week the model produces 12 different betting scenarios. Approximately 6 to 8 of those scenarios can be selected and used as options, providing players enough options to bet on and potentially create parlays.

For the sake of potentially using the model and algorithm in the casino I won’t be releasing the actual nuts and bolts that go into them. Suffice to say they both use proprietary metrics and statistics that I have formulated from the NSFL and both have an acceptable fit to the data.

[div align=\\\"center\\\"]Testing the Model[/div]
I decided to test the model during Week 1 of the S22 NSFL Season. I chose three matchups that the model has the ability to craft betting setting scenario for and put the point spread and over/under for each of those matchups into a google form. The community responded and selected the bet they’d feel comfortable making in a vacuum.

“In a vacuum” is in reference to the fact that there are two levers the casino has the ability to pull to make a bet enticing. The first is adjusting the scenario – i.e. the point spread or over/under – and the second is the adjusting the odds. I did not include odds in the form because I wanted to see which side would get the most action if odds weren’t being considered.

The matchups I chose, scenarios for each, and community results were:

[div align=\\\"center\\\"]Yellowknife Wraiths at Colorado Yeti
Over/Under 60 Points

Yellowknife Wraiths +5.5 / Colorado Yeti -5.5


Chicago Butchers at Philadelphia Liberty
Over/Under 44 Points

Chicago Butchers -6.5 / Philadelphia Liberty +6.5


San Jose Sabercats at New Orleans Second Line
Over/Under 39 Points

San Jose Sabercats +13.5 / New Orleans Second Line -13.5
[/div]

The most popular over/under bet was the matchup between the Sabercats and Second Line. 84.2% of respondents took the Over in the matchup.

The most popular point spread bet was the Liberty at +6.5 in their matchup with the Chicago Butchers. Considering the model is currently only factoring in data from last season and cannot account for the loss of Leaf and other players - and that I didn’t smooth out the line to account for those changes - it is not surprising this was the most popular line.

I then went and ran the algorithm to produce lines for each of the scenarios. As I discussed earlier, both the algorithm and model will get better as I get more data. Some of the odds are going to be a tad bit skewed and would not be the odds that would get used in the casino if this model was being used for that.

[div align=\\\"center\\\"]Yellowknife Wraiths at Colorado Yeti
Over/Under 60 Points
Over +750 i.e. a $250k bet makes $1.875m in profit.
Under -750 i.e. a $250k bet makes $33k in profit
Yellowknife Wraiths +5.5 / Colorado Yeti -5.5
Wraiths +5.5 -400 i.e. a $250k bet makes $62.5k in profit
Yeti -5.5 +125 i.e. a $250k bet makes $312.5k in profit

Chicago Butchers at Philadelphia Liberty
Over/Under 44 Points
Over -350 i.e. a $250k bet makes $71k in profit
Under +120 i.e. a $250k bet makes $300k in profit
Chicago Butchers -6.5 / Philadelphia Liberty +6.5
Butchers -6.5 +500 i.e. a $250k bet makes $1.25m in profit
Liberty +6.5 -500 i.e. a $250k bet makes $50k in profit

San Jose Sabercats at New Orleans Second Line
Over/Under 39 Points
Over -1500 i.e. a $250k bet makes $17k in profit
Under +250 i.e. a $250k bet makes $625k in profit
San Jose Sabercats +13.5 / New Orleans Second Line -13.5
Sabercats +13.5 -350 i.e. a $250k bet makes $71k in profit
Second Line -13.5 +120 i.e. a $250k bet makes $300k in profit[/div]

Some of these lines are not great and probably wouldn’t be used in the event this model was used in the actual casino.

[div align=\\\"center\\\"]Results[/div]
To test how much money was won and lost in the casino I had to assume three things. The first was that the odds placed on these games wouldn’t change the way the community had indicated they’d lean in each scenario drastically, if at all. The second was that the only value bet that could be placed is $250k. The third was that every person who responded would make every bet in a vacuum.

In 5 out of the 6 scenarios there are 38 responses. In the sixth – San Jose at New Orleans point spread - there were 36 responses. This means there is $9.5 million on the line in first 5 scenarios and $9 million on the line in the sixth scenario. The total that would have been wagered this week was $56.5 million.

Here’s how the bets worked out:
[div align=\\\"center\\\"]
Yellowknife Wraiths at Colorado Yeti – Yeti won 34-12
Over/Under 60 Points – Total of 46 points scored; under
31 out of 38 respondents picked the under. Total payout: $8.773 million
Yellowknife Wraiths +5.5 / Colorado Yeti -5.5 – Margin of 22 in a Yeti win.
20 out of 38 respondents expected the Yeti to cover the point spread. Total payout: $11.25 million

Chicago Butchers at Philadelphia Liberty – Liberty won 51-10
Over/Under 44 Points – Total of 61 points; over
24 out of 38 respondents picked the over. Total payout: $7.714 million
Chicago Butchers -6.5 / Philadelphia Liberty +6.5 – Margin of 41 in a Liberty win
32 out of 38 respondents picked the Liberty. Total payout: $9.6 million

San Jose Sabercats at New Orleans Second Line – New Orleans won 23-14
Over/Under 39 Points – Total of 37 points scored; under
6 out of 38 respondents picked the under. Total payout: $5.25 million
San Jose Sabercats +13.5 / New Orleans Second Line -13.5 – Margin of 9 in New Orleans win
17 out of 38 respondents took the Sabercats at +13.5. Total payout: $5.464 million

Total bet: $56.5 million
Total paid out: $48.051 million
Total Casino Revenue: $8.449 million[/div]

The casino operated at a net profit in the scenarios presented in this example. While doing that, however, bettors still had a strong chance to win money. That should be the goal - options where you can still win money if you're interested in partaking, that return some money to the league office so that we don't continue to dilute the value of money in the league, and that are enticing enough for players to want to bet on.

It is important to remember that I picked 6 scenarios out of a possible 12 without knowing the spread, over/under, or odds for any of those 6. On a normal week when there are a full 12 scenarios to pick from the casino could be more nitpicky about which potential scenarios they listed for potential bets. The odds can also be tinkered with or dampened a bit if necessary. This was a pretty extreme set of odds and it still ran at a profit. If this was an implemented system I think it would be necessary to refine the algorithm for the odds so they aren’t as swingy. For example, for this week the casino could have selected the Wraiths at Yeti point spread, the Butchers at Liberty over/under, and the Sabercats at Second Line point spread. Those felt like the most balanced scenarios going in and the community obviously felt the same way when looking at the data.

The model will also grow more accurate and refined as the season goes on. It will begin to put more a weight on this season, meaning situations like a very wonky spread in the Chicago versus Philadelphia game would be less likely to occur. Next week's outputs will already begin to be vastly improved over this weeks.

[div align=\\\"center\\\"]Conclusion [/div]
I think that the way the casino is set up currently is a good start, however it isn’t fully maximizing its potential. Utilizing this model, one that I’m still calibrating myself and will also improve on its own as it gets more data, to get a starting point for spreads, over/unders, and odds would allow the casino to become a more fleshed out entity that would give players a better experience. Although this study makes a lot of tentative assumptions most of the kinks could be worked out in a matter of weeks. This experiment still shows the model and system it relies on has potential for use in the casino.

The work necessary to run a system like I presented today is minimal. I can maintain the model on my own for the most part. A team could be created to process the bets on a consistent basis. It wouldn’t be hard to keep it efficient and would also open more league jobs for potential new players to get involved in the management and future of the league.

[Jay_Doctor]

Sounds promising!

[Vorshayla]

As i've stated in the feedback forum, the only options right now are prop bets.. and not very good ones with the return being atrocious.

If you want to operate a sportsbook, that's fine, but then you need to keep tally of all bets, what your initial line was, and whether you're going to move that line for sharp players.

For example,

If you are betting the spread and your initial offering is extremely likely to have one outcome over the next, people are going to bet on that outcome at say. -750. Eventually you're going to have to move that line to -800 or even -900 so you hedge your bets as the casino.

And generally speaking, when sportsbooks do this.. players don't see the line changing because if you were one of the few who got in when the line was -750, you're going to earn more than the poor sap who was late and got in at -900 and shits gonna hit the fan. But in the real world, lines change based on betting patterns

I've offered to help make it more enticing but no response as of yet

[iStegosauruz]

As i've stated in the feedback forum, the only options right now are prop bets.. and not very good ones with the return being atrocious. 

If you want to operate a sportsbook, that's fine, but then you need to keep tally of all bets, what your initial line was, and whether you're going to move that line for sharp players.

For example,

If you are betting the spread and your initial offering is extremely likely to have one outcome over the next, people are going to bet on that outcome at say. -750.  Eventually you're going to have to move that line to -800 or even -900 so you hedge your bets as the casino. 

And generally speaking, when sportsbooks do this.. players don't see the line changing because if you were one of the few who got in when the line was -750, you're going to earn more than the poor sap who was late and got in at -900 and shits gonna hit the fan.  But in the real world, lines change based on betting patterns

I've offered to help make it more enticing but no response as of yet

Yeah I understand what you're saying. I'm operating in a world where lines don't move because for the purpose of the experiment it would have been too complicated to include. Thats definitely something that should be factored in to whatever calculus HO decides to use with how they operate the casino. In an ideal world the lines should move - I just can't build it into a static study.

I definitely think they need to start accepting community help for the casino though. As it stands I think there is very little reason to participate besides as a minor gimmick. It has lots of potential, it just needs to be enhanced a little.

[Vorshayla]

Yeah I understand what you're saying. I'm operating in a world where lines don't move because for the purpose of the experiment it would have been to complicated to include. Thats definitely something that should be factored in to whatever calculus HO decides to use with how they operate the casino. In an ideal world the lines should move - I just can't build it into a static study.

I mean if lines don't move ever, the only sensible thing would be to make all payouts even money. Anything else would skew it considerably. Since there's no tangible stats to track and sim gonna sim, so it's gonna be hard to even draw the initial line and if you start doing crazy shit like offering arbitrary ratios then the people who know what betting is all about is just gonna play the lines and not the actual bet itself since the payout is gonna be so much better.

I'm going to assume the purpose of the sportsbook is to take money out of the league, so you definitely don't want to have the opposite effect with inflated lines.

For example :

San Jose Sabercats https://i.imgur.com/lZGDADN.png at New Orleans Second Line https://imgur.com/ez5u8bq.png
Over/Under 39 Points
Over -1500 i.e. a $250k bet makes $17k in profit
Under +250 i.e. a $250k bet makes $625k in profit
San Jose Sabercats +13.5 / New Orleans Second Line -13.5
Sabercats +13.5 -350 i.e. a $250k bet makes $71k in profit
Second Line -13.5 +120 i.e. a $250k bet makes $300k in profit



The cats being favored to cover at -350 is a steal. but you're also giving them a huge spread. I would have lowered either the point spread and kept the ratio the same, or increased the ratio and kept the point spread the same

like NO being the underdog in that bet scenario and having to win by 2 touchdowns is a lose lose bet for anyone. No one in their right mind would take that bet because they're the underdog AND they have to cover the spread

[iStegosauruz]

I mean if lines don't move ever, the only sensible thing would be to make all payouts even money.  Anything else would skew it considerably.  Since there's no tangible stats to track and sim gonna sim, so it's gonna be hard to even draw the initial line and if you start doing crazy shit like offering arbitrary ratios then the people who know what betting is all about is just gonna play the lines and not the actual bet itself since the payout is gonna be so much better. 

I'm going to assume the purpose of the sportsbook is to take money out of the league, so you definitely don't want to have the opposite effect with inflated lines.

I agree. I think the lines should move. Whether HO decides to run it that way or implement any of the ideas people have is something I've got no input or information on. I was just talking about specifically I couldn't have them move in the way I conducted testing the model. Having them move would only smooth the model in actual practice - hence why it would be beneficial/necessary in a world where something like this get implemented. I was just hedging that I have no idea what the plan is.

[AdamS]

This season is primarily a testing season for the concept before expanding it to include more options. A bank sheet is incoming and plans are in place on that end.

I did get nearly 2 days though before someone complained though so that's gotta be a record.

[iStegosauruz]

This season is primarily a testing season for the concept before expanding it to include more options. A bank sheet is incoming and plans are in place on that end. 

I did get nearly 2 days though before someone complained though so that's gotta be a record.

Not sure what you're referencing in terms of complaining, but if you meant this post it wasn't mean to be one. You know I've reached out to you personally to give my thoughts on the casino and that I talked through everything in this post with you already. This article was more meant to be a conversation starter for things the casino could be. I don't think there's anything inherently wrong with people giving their feedback and its sparked a pretty interesting discussion between @Vorshayla and I about the methodology behind constructing lines. Posts like these that spark conversations are great ways for y'all in HO to get honest assessments of endeavors from people who care about their outcomes enough to put together projects like this.

[AdamS]

Not sure what you're referencing in terms of complaining, but if you meant this post it wasn't mean to be one. You know I've reached out to you personally to give my thoughts on the casino and that I talked through everything in this post with you already. This article was more meant to be a conversation starter for things the casino could be. I don't think there's anything inherently wrong with people giving their feedback and its sparked a pretty interesting discussion between @Vorshayla and I about the methodology behind constructing lines. Posts like these that spark conversations are great ways for y'all in HO to get honest assessments of endeavors from people who care about their outcomes enough to put together projects like this.

The joke was referring to the feedback thread. I'm actively engaged in feedback.

[iStegosauruz]

The joke was referring to the feedback thread. I'm actively engaged in feedback.

Oh okay awesome I just really didn't mean to offend you if you thought this was meant as a complaint.