How to Interpret the Model
When you find yourself on the Esports Forecast model page, all these numbers may look intimidating at first. Feel free to click around and hover over things — you’re not going to break anything! Let’s take a look at what info is provided so that you can get the most out of our model forecasts.
First, note that wherever you see Rating — this is our model’s proprietary rating for that team or player. It is an indicator of that team (or player’s) strength. A higher rated team will be a favorite against a lower rated team. High ratings will be orange and low ratings are blue — an indicator of whether the team is hot or cold.
If you see PROV listed for a rating — this means the rating is provisional, meaning that the model does not yet have enough data on this team to provide an accurate rating.
Upcoming Series Page
This view is hopefully self-explanatory for the most part. You have a listing of upcoming series, arranged chronologically, in your time zone. Then you can see a short form of the event name — hover over this for the full event name. You can then see the model rating for each team listed.
In the center of the view, you will see how many maps are in the series, Bo1 means best of one, Bo3 means best of 3, etc. If there have been any maps played already, this will show the current score.
Next to that, in the W% column you’ll see the model’s predicted chance for each team to take the series. This will be as of the start of the series — even if some maps have already been played.
The Show PROV button will reveal series with provisional ratings — the model cannot provide an accurate forecast for these series, so they are hidden by default.
Recent Series Page
The recent series view has everything shown in the upcoming series view, but has a bit more colour and information.
In this view you’ll see the results of the series — the winning team will be green and losing team will be red. You can quickly compare this to the colour of the model’s win probability to see how the model has performed over these series. In the top right, we have a measure of the model’s accuracy over all the recent series shown.
Model Forecast Page
When you click on a series, you’ll enter the model forecast page for that series. The first thing you’ll see is an overview for the forecast, showing the teams, the model’s rating for each, and the model’s predicted win percentage for each team to take the series.
If maps have been played, you’ll see the current series score. The icon under the score indicates the esport being played — we’ll look at a Dota 2 series here as an example.
The next table that you see shows the model’s rating and win probability for each map in the series. You can click on each map here to update the entire page to see the forecast as of the start of that map.
Below this table you’ll see esport specific factors that the model is taking into accounting. This updates map-by-map.
In this example we have map 1 selected (by default) and it shows that the model gives a small advantage (+5) to Team Liquid (on the right) because they have first pick. The model does not detect a Radiant/Dire bias at the time when this series was played.
For esports with map picks (Counterstrike, Overwatch), you can input the selected maps using this table. If you are watching a stream, you may get this information faster than it becomes available within the model, so adding in this factor manually will provide you with a more accurate forecast.
Next you will find the expected rosters for each team.
It is recommended to double check these rosters with a secondary source to ensure that the model forecast is accurate. If you see any issues with rosters, you can let us know in our discord, and we will correct the issue.
Next, you will find the table of value bets that the model calculates. This table contains a variety of popular bets that you can find offered by most bookmakers.
In the center of the table you’ll see the name of the bet that we are looking at — at the top here we see the bet for which team will win the series outright.
The % columns are is the probability with which the model expects each outcome to occur. So looking at the top row here — the model calculates the left team to win the series 38% of the time, and the right team win 62% of the time.
The Min.Odds columns show the minimum recommended odds for which you should consider the bet. If you can find these odds or higher at a bookmaker, placing this bet would have positive expected value according to the model — this is a value bet.
In the top left of the table you can switch between Decimal or American odds, depending on your preference.
If you decide to place a bet, you can click on the % or Min.Odds column for that bet to enter it into the Kelly Bet Sizer table that follows — we’ll consider a bet on Team Liquid to win the series.
This table will help to decide how much to wager using the Kelly Criterion — a popular method of bet sizing among sportsbettors. The cursor is immediately placed into the Bookie Odds column so that you can quickly enter the odds available at your bookmaker. You can also set a Bankroll and Kelly Fraction, which the browser should remember for your next visit.
You’ll see the Edge of the bet, which is a measure of how profitable the bet. Most importantly, you’ll then see the optimal Kelly bet size in the Bet Size column, and the expected value of the bet in the final column.
Hopefully this clears everything up about the model forecast and series overview pages — if you have any questions, please leave a comment.