Remi's Data Methodology
We Get It - Numbers Can Be Overwhelming
To simplify, here's what each number & term means and how it's used in our calculations.
See these numbers in action:
Common Sports Betting Numbers & Meanings
Spread bets are wagers placed against a spread line. The spread line is a sports betting number made by oddsmakers to handicap the two teams against one another. When someone places a spread bet, they are betting their team PLUS the spread line will beat the other team. Note the spread line can be negative, which means that team is favored.
A pick of TmA -4 means TmA must win by at least 4 points to cover the spread. A pick of TmB +4 means TmB must win, or lose by less than 4 points to cover the spread.
Another way is a pick of TmA -4 means (TmA -4) must be greater than (TmB) to cover the spread. A pick of TmB +4 means (TmB +4) must be greater than (TmA) to cover the spread.
Moneyline bets are wagers placed on a game’s outcome. They are bets on which team or competitor will win any given match, and the winner will receive a payout at the odds the bet was accepted. When Remi releases a moneyline pick, it will shows in American odds (-110, +100).
Total or Over/Under bets are bets placed on how many points will be scored in a game. For example, if oddsmakers set an over/under line of 212, and if the total points scored in the game is 200, the game went ‘under’, and any bet on the under for the game would win the payout odds the bet was accepted at.
American Odds are way of presenting sportsbook odds that show how much money must be wagered to win $100 in return (for negative odds), or how much one would win if $100 was wagered (for positive odds). For example, odds of -300 mean one would have to bet $300 to win $100 if their team won. Odds of +140 means a wager of $100 would win $140 in return.
Decimal Odds are way of presenting sportsbook odds that show how much money will be won if the bet wins as multiplied by a single number. For example, odds of 1.91 mean if one bet $100, they would win $100 * 1.91 = $191 if the bet won. Once the original wager amount is subtracted, they would generate $91 in profit.
A pick is when Remi finds sufficient evidence to calculate a positive Expected Value (EV), or positive ‘Units’, regarding a team to win outright (Moneyline), cover a spread (Spread), or a game to go over/under the line (Total).
When the outcome Remi picked happened in the game, and bets on those sides of the line won.
When the outcome Remi picked did not happen in the game, and bets on those sides of the line lost.
When the outcome of the game was exactly the same as the Average Vegas Line, and the play is cancelled.
A Unit is calculated on a mathematical formula (we use a modified Kelly Criterion approach) correlating expected win probability and payout odds. This is also referred to ‘EV’, but for most purpose we’ll use a ‘Unit’.
Typically 1 unit applies to 1% of bankroll, however some find this too aggressive (since Remi can release 20-30 units each day), so often times a ‘Quarter-Kelly’ (1 unit = 0.25% of bankroll) or a ‘Half-Kelly’ (1 unit = 0.5% of bankroll is used).
Just a reminder, this is only for informational purposes and is NOT a recommendation as to whether or how anyone should wager.
The strength of the predicted win probability vs the implied probability, strong positive EV, or strong positive ‘Units’.
For most spread picks, the implied win probability is 50%, and after vigorish this changes to 52.4%. In this case, a lean would be any projected win probability greater than 52.4%, and the higher the win probability, the stronger the lean
Wins against losses, such as (521-412)
Wins divided by total games, such as (521/933 = 55.8%)
Wins minus losses, such as (521 – 412 = +109)
The cut a sportsbook takes on a bet. For most bets Against the Spread, this is 1.91, or -110 (meaning it takes $110 to win $100).
A no-juice or no-vig market would mean spread bets would have +100 or 2.0 odds, but in order to stay in business, the sportsbooks take 9% of every winning bet and keep it.
Remi Specific Numbers & Meanings
Remi uses an average line across most sportsbooks. Since lines can fluctuate from region to region, sportsbook to sportsbook, and change throughout the day, Remi picks an average line when picks are released to run probability calculations against. For calculations below, we’ll call this the ‘Average Vegas Line’.
Remi uses average odds across most sportsbooks. Since odds can fluctuate from region to region, sportsbook to sportsbook, and change throughout the day, Remi picks average odds to run unit calculations with. For calculations below, we’ll call this the ‘Average Vegas Odds’.
Remi’s data is tracked back to September 1, 2022. While the tech that feeds Remi has been in development for years, this is when we started tracking Remi’s performance.
In December of 2022 we conducted a data migration effort between two web servers. During this process, several small segments of picks between Sept 1, 2022 and Dec 15, 2022 were lost for certain sports, and as a result we decided to part ways with our web migration company. As a measure of good faith, any metrics using this 4-month period will still use the full value for ‘Remi’s Fees’ (-$299/mo) and will not deprecate the fees for some ratio of the lost data, even if Remi might have done better than reflected in the metrics presented.
When Remi releases a pick, this is the probability that team will win outright (moneyline), the probability a team will cover the spread (spread), or the probability the game will go over or under the Average Vegas Line (total).
This is denoted as %WIN or PROB in certain tables
Remi’s data across every sport Remi covers, including NFL, CFB, NHL, MLB, CBB, and NBA.
Metrics involving Remi’s record picking against the spread and over/under.
Since 99% of Remi’s picks include odds between -130 and +120, we’ve also included near toss-up moneyline picks into the against the spread metrics.
Except in extremely rare circumstances where the Units or EV say otherwise, Remi avoids favorites stronger than -130.
A flat -110 vigorish (vig or juice) is applied to every win being analyzed.
Note: Slight favorites (-120) and slight underdogs (+120) are also calculated using the flat -110 vigorish, meaning a slight favorite will have a slightly worse payout than the flat -110 vigorish, and an underdog will have a slightly better payout than the flat -110 vigorish.
We’ve found the -110 flag vigorish is usually within 2% percent of reality either way since 99% of Remi’s picks are between -130 and +120 odds. We’ve also found that over the long term, most of the slight favorites and slight underdogs average themselves out to near a -110 vigorish.
Finally, we’ve chosen this because payout odds can vary from region to region and sportsbook to sportsbook, so the flat -110 vigorish is more fair than showing partiality to odds in certain regions or certain sportsbooks that not everyone may have access to.
Remi’s winning units minus losing units, minus a flat -110, or 9% vigorish (juice) on every win. See why we use a flat -110 vigorish in the ‘After Vigorish’ pane.
he net units after a flat -110 vigorish over the last 20 games for all 6 sports combined (totaling 120 picks). See why we use a flat -110 vigorish in the ‘After Vigorish’ pane.
An ESTIMATION of past profit using a flat -110 vigorish, Remi’s monthly fees of $299 over the duration being analyzed, and using a certain dollar/unit correlation.
Since Remi uses a flat -110 vigorish on every pick, this is just an approximation. See why we use a flat -110 vigorish in the ‘After Vigorish’ pane.
Please note, these metrics are trailing indicators and NOT predictive indicators. Remi is a machine learning algorithm that is constantly changing and learning for better or for worse, so past performance and profit estimations are in no way indicative or predictive of future results.
Aggregate statistics for every pick released of a certain unit value. For example, every ‘5’ unit pick released has won 55.1% of the time over 436 picks.
Floor calculation for a unit value and every unit value higher. For example, 7+ means every ‘7’, ‘8’, ‘9’, and ’10’ or higher unit pick.
Used in profit estimation metrics based on a dollar figure per unit played. This is for informational purposes only as a way to visualize past data, and is NOT a recommendation as to whether or how anyone should wager.
How well Remi performed as a win percentage since Remi’s tracking began Sept 1, 2022. Read more on the ‘How Long Has Remi Been Around’ tab about the data migration effort of the fall of 2022.
A mathematically generated probability using a modified Kelly formula and payout odds that correlates a win probability to a unit value. For instance, a ‘4’ unit play is expected to win ~54% of the time at -110 odds.
Remi’s monthly fees calculated at $299/mo over the duration of the analysis.
Note, we do not include a free trial month in fees calculations for metrics. If the window in question goes back 3 months, the fees will be $299*3=$897 and not $299*2=$598 even if clients got a month free when coming onboard.
The data presented in tables and past record are TRAILING INDICATORS and NOT PREDICTIVE INDICATORS. Remi is a machine learning algorithm that is constantly changing and learning for better or for worse, so past performance and profit estimations are in no way indicative or predictive of future results.
Past profit metrics are ESTIMATIONS ONLY using a hypothetical dollar to unit amount and a flat -110 vigorish approximation. In no way does the platform, Leans.AI, or the picks generated by Leans.AI constitute financial advice. The data presented is for informational purposes only and is not a recommendation as to whether or how anyone should wager.
Remi’s data is tracked back to September 1, 2022. While the tech that feeds Remi has been in development for years, this is when we started tracking Remi’s performance.
In December of 2022 we conducted a data migration effort between two web servers. During this process, several small segments of picks between Sept 1, 2022 and Dec 15, 2022 were lost for certain sports, and as a result we decided to part ways with our web migration company. As a measure of good faith, any metrics using this 4-month period will still use the full value for ‘Remi’s Fees’ (-$299/mo) and will not deprecate the fees for some ratio of the lost data, even if Remi might have done better than reflected in the metrics presented.