Deeper Betting Knowledge Part 2: Using Monte Carlo Simulations To Evaluate Tipsters

Welcome back to the 2nd part of our look at the 12 Advanced Tipster Outputs we now publish in every SBC Tipster Review and how they can help reveal the best, most reliable tipsters.

If you missed Part 1, you can read it on the SBC Blog via this link.

Otherwise strap in for outputs 5 to 12, what they are and what they tell us about a tipster service…

5) 50th Percentile Drawdown

This is obtained by the same method as that in the 99th percentile drawdown (#4 in part 1), although on this occasion, we look at the 50th percentile, This calculation indicates there is a 50% chance, i.e. even money, of a losing run in any given year to the current staking plan.

It indicates that, if a service is followed for two years, a bettor should absolutely expect to suffer a drawdown of this size at some point.

What To Look For: A figure that is absorbed by the suggested betting bank as we can expect this run to happen once every 2 years. Again, it’s useful for you as a punter as you might want to avoid services with big drawdowns likely to occur at least once every 2 years.

Recent Examples: One tipster provided quite a severe example of this in action as we discovered that their 50% percentile drawdown figure was a staggering 111% of the suggested betting bank (as put forward by the tipster themselves). Clearly this was inadequate and if following the tipsters bank suggestion, you should expect to go bust once every 2 years. It was no surprise this tipster did not receive a strong rating from the SBC team.

6) 50% Bankroll Drawdown

Description: The result of this calculation, also known as the ‘semi-ruination’ figure, indicates how often one is likely to avoid a drop that eats half of the funds set aside for betting with a service.

It is a widely-held belief that some considerable resolve and confidence is required once you lose half of your betting bank. At the point that 50% of your bank goes, most people would quit.

What To Look For: The higher the percentile that this output gives, the better and the less likely one is to have to contend with such a 50% dip.

Recent Examples: One extreme example of a service with a poor 50% bankroll drawdown figure came in a review of a racing tipster from 2019, which indicated you should expect to see a peak-to-trough drop of this size in 82% of all years. Effectively meaning that every 4 out of 5 years you would lose half your bank and helping to highlight a very volatile service.

7) Likelihood of a Losing Year

Description: A percentage figure that indicates how likely you are to suffer a losing year.

What To Look For: The lower the figure the better – especially for those of you who expect to play it ‘safe’.

Recent Examples: The extensive simulations that we have run during the last year show one top rated SBC Hall of Fame racing tipster to have the lowest probability of a losing year. In fact, our Monte Carlo simulator ran 5000 seasons covering almost 28m bets and not one losing year was encountered!

The highest probability recorded was just over 49% by a poorly rated racing service. This meant that a loss should be expected every other year if following and it is of little surprise that this service ended a few months after our analysis.

8) Risk Reward Ratio

Description: This ratio is obtained by taking the average annual profit divided by the average annual drawdown over the period of the review.

What To Look For: The higher the number the better and for no huge disparity between the live and simulation results.

Recent Examples: Anything above a score of 2 can be regarded in a positive light (classed as ‘excellent’) and several recent tipsters all achieve this with a very high score for one racing service who hit a mark of 3.427..

9) Capital Risk Ratio

Description: The Capital Risk ratio is a simple calculation that represents the percentage of the bankroll consumed by the maximum drawdown suffered.

What To Look For: The lower the number the better and again parallels between the tipster’s live results and simulation results are welcome.

Recent Examples: If a tipster’s largest drop is less than one-third (33%) of their recommended bank, then this wins a rating of ‘excellent’. A score of more than 66% is a concern and noted as ‘poor’ and, again, intermediate figures are labelled as ‘strong’ and ‘average’.

A recent NFL tipster we reviewed leads the way in this exercise with a ratio of 17.20% with several others all featuring around the 22% mark. These all constitute an excellent Capital Risk Ratio score.

10) Dispersion Factor

Description: The dispersion factor indicates the degree of instability or uncertainty that should be expected when following a tipster. We use the simulations to produce best and worse-case strike-rates, disregarding the top and bottom 5% of results.The dispersion factor is the ‘near-best’ figure divided by the ‘near worst’, i.e. taking the 95th and 5th percentiles.

This output is routinely used to remove the outliers or freak results such as a one-off 200/1 winner or others that vary from the norm.

What To Look For: The lower the factor, the less volatile the tipster should be.

Recent Examples: A recent SBC review tackled a tipster with an excellent ROI of over 30%, but as one targeting bets at over 10/1, it can be a volatile service to follow. This is reflected by the fact they have a dispersion factor rating of 1.796, which is rated as ‘high’.

This follows the accepted wisdom that any investment with a high reward, for example a ROI of over 30%, comes at a high risk.

11) Sharpe Ratio

Description: This is a relatively new index that we have introduced into our analysis work to provide a measure with which we can compare a tipster’s annual betting bank growth.

It compares this growth against the UK’s average ‘risk-free’ investment rate over the review period. Effectively – what kind of interest rate you might enjoy if putting your money in guaranteed sources such as a bank savings account or a government bond. The risk-free rate has been consistent over the last few years at around 2.1% (although may admittedly be changing in this current climate). The ratio is the average ROC returned in excess of this figure.

What To Look For: The higher the number the better. It is also a good examination of what you can make in comparison to other savings options out there.

12) P-Value

Description: An incredibly useful output, the p-value is a statistical test to evaluate the probability of obtaining a set of results by chance (as opposed to skill). It is obtained from an algorithm that uses three variables: overall strike rate, ROI and the average odds of all selections.

What To Look For: The lower the number the better and as close to 0 as possible. A high number indicates either a lack of live data or a set of results that might be based on luck.

Recent Examples: ‘Low’ probability covers the range between 0 and 0.025, whilst a result of between 0.025 and 0.05 is seen as ‘moderate’, with p-values in excess of this being taken as ‘high’. To complete the picture, a p-value of 1 would indicate absolute certainty that the results have been obtained by pure chance alone.

Several strongly rated tipsters we have reviewed recently had p-values of zero, which suggest that the results obtained were 100% based on skill and there is no luck element to be found. There are also numerous other tipsters that have gone under our result microscope, each of whom had very low p-value scores.

Those with higher p-value scores are usually those tipsters either targeting bigger priced selections OR those with smaller data sets than we would like. For example, one promising golf tipster had a high p-value of 0.3130 as he had put up less than 600 tips over 3 years and at a strike-rate of 20.54%. Highlighting this as a tipster or promise, yet one we clearly need more data for before rushing to invest our money in his advice.

High scores could, but not necessarily, indicate that a tipster has changed methodology of selection during the review period and this is something we would test with the tipster.

Advanced Tipster Analytics Wrap

I do hope you have enjoyed this special 2-part guide on these Advanced Tipster Outputs and just how they can help you discover more about betting services.

They go further than simply evaluating which tipsters are good or not, but into other key areas such as losing runs, bank sizes and the level of risk if following them in.

All these outputs are important so you can fully understand which tipsters you should follow and those that suit your personality.

If you want to know more, then you can explore the full Advanced Tipster Analytics Guide with a Smart Betting Club membership. Best of all – you can read the 12 scores and ratings for each tipster we review based on the Monte Carlo Simulations we perform as standard.

All of which are designed to give you every bit of information to help you choose the best tipster services for you!

SBC Helping Inform Your Betting Over The Next Few Months

Whilst there isn’t a huge amount for us to bet on right now, we are using this quiet betting period to publish a series of quality reviews, articles and insightful guides to help deepen the expertise of SBC members.

Mindful of the fact that many of you are using this time without sport to improve your knowledge and understanding of the betting and tipster world, our role is to help fill that gap with informative, educational and quality content.

Tackling everything from helping develop and understand your betting risk profile, the psychology of a winner, Advanced Tipster Analytics and guides to betting on new sports, you can expect to read a lot as an SBC member over the next few months.

As ever, membership comes with a full 30 or 90 day money back guarantee, so if you are looking to develop your betting skills, do consider a Smart Betting Club subscription.

Best regards

Peter Ling

Smart Betting Club Owner and Founder

 

 

Deeper Betting Knowledge Part 1: A Guide to Using Monte Carlo Simulations To Evaluate Tipsters

In these quiet times for sport and betting, it’s an ideal time to look to develop our knowledge and skill sets and as part of that, in this new 2-part series I want to walk you through how to use Monte Carlo simulations to evaluate tipsters.

Much of this work has been ported over from other investment worlds and the outputs provided help us make more informed decisions on just which tipsters offer the best chance to make you a profit betting.

The full SBC Advanced Tipster Analytics Guide on Monte Carlo Simulations and the outputs for each tipster are available to full SBC members, yet in this 2-part article, I want to outline what they are and how exactly they can help when evaluating a tipster.

The good news is that you DO NOT need to understand any complicated mathematics to benefit from exploring these advanced outputs. Simply follow the easy to understand explanations we give and the context and scores we put the ratings in.

These advanced analytics are ideal if you are interested in topics like losing runs, risk, whether a tipster is lucky or skilful, comparison to other investment returns and volatility.

Over the course of this 2-part article as well as explaining the usage of Monte Carlo simulations behind the analytics, I will also be explaining the outputs they provide, which include:

  • Historical maximum drawdowns
  • 99th and 50th percentile drawdowns
  • 50% bankroll drawdowns
  • Likelihood of a losing year
  • Risk reward & Capital risk ratios
  • Dispersion factor
  • Sharpe ratio
  • P-values

If this all sounds complex – don’t worry it isn’t and please don’t stop reading!! You honestly don’t need a degree in maths to figure out how they can help you.

Instead let me guide you through each of them and explain what they are and how they can help you make better decisions for choosing the tipsters you follow.

Introducing Monte Carlo Simulations

When the SBC team reviews a tipster, the first port of call is to explore the live record of performance obtained to date. Usually this encompasses at least a couple of years data, if not more and often at least 1000 to 2000 past actual bets.

As useful as it is to explore this live actual record of performance, from an analysis perspective, the amount of data we have – even if 2000 bets is often insufficient to draw 100% concrete conclusions from.

This is where running Monte Carlo simulations for each tipster comes in as they replicate the profile of the service over many million data points.

So rather than calculate what might happen when analysing 1000 or 2000 live bets, we do so over say 90 million – a much more useful data set.

The results of each Monte Carlo simulation can help to answer key questions for any given tipster such as:

  • Is a betting record more likely to be based on luck or skill?
  • What kind of losing runs or drawdowns might you realistically expect?
  • How do their live results compare to their simulation results?
  • Have they benefited from some freak results and outliers that are unlikely to be replicated?
  • What size betting bank do you REALLY need?

Ultimately, all the outputs from these simulations helps to give us a greater understanding of the quality of any given tipster. Certainly, it is far more useful than just scrutinizing the live results only.

All SBC reviews now include the results from our Monte Carlo simulator outputs as standard, hence why I wanted to illustrate what they mean and how you can interpret them in this special 2 part article.

How These Simulations Work

Monte Carlo simulations have existed in several forms for around 90 years and have been used in many walks of life such as medicine, insurance, space, oil exploration, nuclear weapon experimentation and even for general election modelling!

The simulation is a mathematical technique which is used to understand the impact of risk and uncertainty.

The main principle is the application of randomness and volatility to test theories of probability.

They are absolutely ideal for evaluating historic betting patterns and to predict best and worst-case scenarios.

When we at the SBC take a tipster’s proofed selections and provide statistical analyses by, for instance, season, month, course, race type and odds banding, we augment this by taking several copies of the key data fields and then modelling these through a computer algorithm which generates ‘virtual seasons’ each of the mean number of bets.

Typically, this might create, say, 100,000 seasons which could produce in the region of 90 million bets!

The greater the number of simulations that are run, the more confident and positive one can be of the accuracy of the outcomes.

It often takes 5-6 hours to run these Monte Carlo simulations for each tipster and a fair bit of processing power to get there but the results are often well worth the wait!

12 Outputs to Rate Tipsters By

By running the Monte Carlo simulations against each tipster analysed and reviewed, we are able to produce detailed results that provide a number of very useful outputs – 12 in total.

Over this 2 point guide, I will provide a description of each output with our reasoning as to why they are useful analytical tools and their application, together with some worked examples from our recent investigations showing how you can use them to make better decisions.

1) Strike Rate

This first output is a simple one as the strike rate is simply the overall number of winners divided by the tips given, expressed as a percentage.

What To Look For: Whilst there is isn’t really a ‘good’ or ‘bad’ strike-rate, this output is important as those tipsters with a lower strike-rate often require more patience, yet bring higher rewards and vice versa.

2) Return on Investment (ROI)

Another standard output whereby the ROI indicates the actual profit as a percentage of total stakes. ROI is always useful as it puts profit into context.

What To Look For: The type of ROI you can make often varies based on the strike-rate and the type of sport the tipster specialises in.

It isn’t rare to see a top racing tipster hit 30% ROI+ (caveated by the issues of getting on) vs that from a football tipster, whereby anyone making 5% ROI+ is doing very well (caveated by the fact it’s often much easier to get your bets on).

Therefore, the ROI needs to be judged in context against the markets tackled and the profile of tipster.

3) Historical Maximum Drawdown

Moving onto the first of the more advanced outputs and it’s important to understand what a drawdown is as this is a term we use regularly.

A drawdown is quite simply the worst historical run that has ever taken place – the peak-to-trough decline in bankroll experienced over a period of bets.

What To Look For: A tipster service with a strong record will have a historical maximum drawdown that is easily absorbed by the betting bank suggested. It’s a useful output to examine as if you struggle with long losing runs, you want to avoid tipsters that suffer occasional large drawdowns.

4) 99th Percentile Drawdown

Building upon [3] above, what this output does is effectively list all of the seasonal drawdowns found in the simulations and then takes the 99th percentile figure. It is useful as should a drawdown exceed this 99% figure; it indicates there is a problem or change in the service to be aware of.

What To Look For: We like to see similarity between a tipster’s live record and the results from the Monte Carlo simulation. If a tipster suffers a drawdown that equals or goes beyond this 99% figure – it could suggest a problem.

Recent Examples: We recently analysed a long-term profitable racing system, which over 9 years’ worth of live results had suffered a heaviest drawdown of 109 points. The 99th percentile figure came out at 119 points – suggesting the live tipping history was very similar to the simulation results. A very good sign.

Compare this with our analysis on another racing tipster recently that didn’t attract a high SBC rating. Although their worst drawdown to date was 63 points, the simulations reveal that 139 points would be needed to accommodate the likely potential drop at the 99th percentile. Suggesting that followers of this tipster should prepare for a worse run than that which has been seen in live results.

Read Part 2

Click here to read part 2 of this article, where I cover 8 further outputs from the Monte Carlo simulations including:

  • 50th percentile drawdown
  • 50% bankroll drawdown
  • Likelihood of a losing year
  • Risk reward ratio
  • Capital risk ratio
  • Dispersion factor
  • Sharpe ratio
  • P-values

SBC Helping Inform Your Betting Over The Next Few Months

Whilst there isn’t a huge amount for us to bet on right now, we are using this quiet betting period to publish a series of quality reviews, articles and insightful guides to help deepen the expertise of SBC members.

Mindful of the fact that many of you are using this time without sport to improve your knowledge and understanding of the betting and tipster world, our role is to help fill that gap with informative, educational and quality content.

Tackling everything from helping develop and understand your betting risk profile, the psychology of a winner, Advanced Tipster Analytics and guides to betting on new sports, you can expect to read a lot as an SBC member over the next few months.

As ever, membership comes with a full 30 or 90 day money back guarantee, so if you are looking to develop your betting skills, do consider a Smart Betting Club subscription.

Best regards

Peter Ling

Smart Betting Club Owner and Founder