< View all strategies

XG12 trading strategy

Sharp strategy, clean execution.

Mercurius Tradr is a trading bot for Betfair Exchange that automatically runs our latest trading strategy. The XG12 strategy was released in beginning of 2019 and replaced by the most recent Phoeni Xg in May 2020.

At XG12 core lies an custom extension of a Poisson model - which is the combination of the most performing models produced by the scientific community so far - that captures the relative skills of football teams from Wyscout’s data feed.

1. Sports trading approach

The XG12 approach aims to sports trading is to estimate a probability distribution of future outcomes for a given match which is more accurate than the one derived from the odds available in the betting market. We do this to find operations that yield positive expected value. If you want to learn more about our trading approach have a look at our Methodology page.

2. Betting markets

At the moment, Mercurius’ technology operates on eleven football leagues. The automated strategy is executed within Betfair’s exchange betting on the back side of the “Home-Draw-Away” market. Although the core principles of the strategy are the same for each of the leagues it considers, the triggers for placing orders and the type of outcome it bets on vary. This is the case because our AI confidence in its estimates is different for each league it monitors. Therefore, it needs a higher expected value in leagues where there is a more significant margin of error in the estimates.

On the right: The eleven leagues ordered by degree of positive expected value needed to place an order. The lowest level is at the top.

League Home Draw Away Composition
Premier League 24%
Bundesliga 18%
Eliteserien 14%
Allsvenskan 12%
Chinese Super League 10%
English Championship 7%
Eredivisie 4%
Ligue 1 3%
La Liga 2%
Primeira Liga 2%
Serie A 2%

3. Staking plan & order execution

The staking plan is Kelly’s Criterion capped at 1%. This means every bet will be 1% of a compounded bankroll. We know execution is crucial. Since Tradr handles many accounts simultaneously, and to accommodate all the liquidity it needs to place bets, we have adopted some key features.

Iceberg alghoritm

Iceberg alghoritm

To ensure the desired amount is allocated in the value zone, the algorithm splits the wagers over time to ensure the market is not harmed, and also to reduce slippage.

Clustered execution

Clustered execution

This feature is built to give priority to higher bankrolls, and efficiently place all the liquidity demanded by our clients.

Dynamic value detection

Dynamic value detection

When slippage occurs, if the value moves from one outcome to another, the algorithm will wager the unmatched liquidity on the new outcome as long as it has a positive expected value.

4. Expected performance

Daily profit and loss

Annualized ROC 67.91%
Yield 5.90%
Average odds 4.33
Strike rate 30.6%
Tot bets 4,353
Bets to recover 971

As shown, the performances although yielding an impressive ROC are extremely volatile. As a consequence, to have a statistically significant performance within a confidence interval of 95%, Tradr needs to place around 1,700 bets. This figure translates into a time interval of 3 years, given the state of the tech and the market. When enough bets are placed, the user should expect a yearly return on capital (ROC) of 36%, net of all costs, our fees, and Betfair commissions.

Learn more

Tradr is a fully automated AI-driven bot. It aims to find and exploit inefficient odds offered within betfair.com's betting exchange on football matches. Check out the Performance Page to have an idea of the actual profitability since we officially released Tradr. We also invite you to have a look at our Methodology for a deep-dive into our philosophy.