How to win at sports betting long term.

Bookmakers have strong models and rightfully so, the pricing day to day on thousands of sports events around the world means that their existence as businesses solely depend on how well they can forecast and mitigate/ manage against risk. Entain group posted Net gaming revenue (NGR) figures of 4 billion in 2022 while Flutter posted a group revenue of 7.6 billion. This does not happen down to luck and it is the result of many business functions within these enterprises such as pricing and risk management. Anyone with a slight interest in sports betting knows that the overround (aka the juice,vig,cut,margin) applied to a sports betting margin is the expected amount a bookmaker will make on average. In my experience beating the closing line is the most optimal method of turning negative expected margin (remember the offered odds will always sum up to greater than 100%) into positive is a mathematical approach that will in the long run turn your sports betting positive return.

There are many conventional ways to do this.

1. Become an expert in your sport — this is a heavy time burden and can be fun if you really get into it.
2. Specialise in lower grade sports — where price confidence is not high. Take the English Premier League, its the most liquid league in the world and price confidence is extremely high and it is difficult to beat long term. Table tennis or ITF tennis are considered lower grade sports. Perhaps even a Soccer league in your local country where you have inside information for injuries and team news.
3. Model it — This can be fun but also a heavy financial penalty if your model is wrong more often than not. Usually, this is a good method if you can apply strong sports knowledge to the sport and have a good knowledge of the league it applies to. It is also required to have a strong modeling background.
4. Use code — and this is what im going to propose as a final idea.

An example of a Value Betting System in Python.

Here I will present a working example of my code in action. I have this set up for 5 Sports currently and it works for Euro books. I dont have access to some like 1xBet or 10Bet as they are not available in my country. The VBS tennis script returned several games for today for ITF tennis which is considered lower grade. The game of interest was between Clara Tauson — Elina Svitolina. The code said that a certain bookmaker was out of line from the market average of 5.6%.

Tennis Output from VBS.

Following the link the prematch prices were indeed off with most of the books moving to a closer game than the odds offered.

The odds were indeed the odds advised through the script. Svitolina won the first set and reached a price of 1/14 or 1.07 but subsequently went on to lose 2–1. Perfect example of variance working against here.

However variance in sports betting can be your friend if you have the closing line beat on your side, as the expected returns should move to your favour through repeated experiments. Take this example where I simulate a mythical 50 euro bet on this type of bet. (remember the sport or the selection does not matter just the odds taken.)

I made a dataframe that simulated 10,000 similiar bets to this one. I calculated expected returns based on the 5.6% market average (although this could be done by simply comparing to Pinnacles closing line) and stated what would be the returns if this bet only won 50% of the time even though the market implied around a 53% chance of winning.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt


data = {'Stake': [50] * 10000,
        'Price': [2.1] * 10000,
        'Exp_margin': [0.056] * 10000}

df = pd.DataFrame(data)

df['Exp_return'] = df['Stake'] * df['Exp_margin']
df['Exp_return_incl_stake'] = (df['Stake'] * df['Exp_margin']) + df['Stake']

# Create a new column 'Result' with half 'w' and half 'l' randomly assigned
np.random.seed(1234)  # Set random seed for reproducibility
result_array = np.array(['w'] * 5000 + ['l'] * 5000)  # Create an array with half 'w' and half 'l'
np.random.shuffle(result_array)  # Shuffle the array randomly
df['Result'] = result_array.tolist()  # Convert the array to a list and assign it to the 'Result' column

# Verify that the 'Result' column contains roughly half 'w' and half 'l'
print(df['Result'].value_counts())


# Calculate 'Actual_return' based on the 'Result' column
df['Actual_return'] = np.where(df['Result'] == 'w',
                               df['Stake'] * (df['Price'] - 1),
                               -df['Stake'])

# Add a new column 'Bet_count' that starts at 1 and increments by 1 for each row
df['Bet_count'] = np.arange(len(df)) + 1

df['Cumulative_exp_return'] = df['Exp_return'].cumsum()
# Add a new column 'Cumulative_return' that cumulatively sums the 'Exp_return' values
df['Cumulative_return'] = df['Actual_return'].cumsum()

# Plot 'Bet_count' against 'Cumulative_exp_return' and 'Cumulative_return' on the same graph
plt.figure(figsize=(12, 8))
plt.title('Cumulative Returns')
plt.plot(df['Bet_count'], df['Cumulative_exp_return'], label='Expected')
plt.plot(df['Bet_count'], df['Cumulative_return'], label='Actual')
plt.xlabel('Bet count')
plt.ylabel('Cumulative return')
plt.legend()
plt.show()

It can be observed that taking an positive margin position (blue line) sees the returns trend up in a linear fashion. The actual returns deviate from wins and losses which can be seen in the rise and falls of the line (orange). But what this simulation does show is that the returns do trend upwards overtime. So this is the expected result of such a strategy. I have highlighted expected there to reiterate that variance could result in 5, 10 even 15 losing bets in a row but as the sample continues the result is that this strategy is expected to be profitable.

Expected versus actual returns in the simulation.

Naturally, risk management professionals do not like customers obtaining continual edges and deal accordingly. There are ways of making your account last longer but require a lot of ‘square’ betting to cover the tracks from the risk algorithm. Let me know if you have any ideas or thoughts on this matter at my email footballdotpy@gmail.com