It seems that these days the corona virus has taken over pretty much all the media and our daily lives. Today I would like to offer you some change of that routine with a review of a (relatively) new value betting company called Mercurius and their predictive Mercurius Tradr model identifying value on the football market. I will look into Mercurius’ concept for identifying value, their performance-based fee structure as well as my experience with the service.
This review is probably a bit overdue, as I have informally announced it some months ago and was approached by a few readers about it already. However, it was important to me to gather a bit larger collection of bets in order to be able to draw some conclusion about the quality of the pics. Of course, and as you know, even long profitable series are no guarantee for an edge. Still, a longer profitable series can give you a bit more confidence, so in that context more is always better.
I do not have an affiliate deal with Mercurius, nor do I use the software at any preferential conditions. I am continuously failing to generate any substantial affiliate income from this blog, so I thought I could spare myself the trouble of joining yet another affiliate program. Perhaps a trend that will continue for the future of this blog. That aside, here are my impressions and experience after around half a year of using the Mercurius Tradr value betting software.
Where it all started
In Summer 2019 I got approached by Carlo from Mercurius asking if I would be interested to review their product. I had a quick look at their website, got familiar with the basics and thought the concept had some potential, so I decided to give it a chance. I have set up my account, funded it with GBP 1000 and started collecting a betting record and some first impressions. Here is what Mercurius Tradr offers at the most basic level.
Targeting the European football markets
Mercurius Ltd is a company that develops betting models, which aim to discover value on the football markets in big European leagues. That is quite an ambitious task. Big league football markets attract a lot of money, so you would expect the odds to be a lot sharper that what I am up against with my League of Legends betting model for example. On the other hand, there is probably enough liquidity to be able to turn over a substantial betting volume. This makes the market a fitting target for a bigger, well-organized and well-funded organization. Therefore, the guys form Mercurius decided to try their luck there. Not an easy task for sure, but not impossible either.
The process is designed to be seamless, with little to no effort for the customer. For that, Mercurius have partnered with Betfair where all the bets are placed at 2% commission. Basically, the punter makes his deposit, clicks the Start button and comes back some time later to check on the progress. The ultimate wish of the team was to offer an investment-like product, which allows the customer to generate a passive income. The bets are being placed by a betting bot, requiring zero effort from the customer. That income is than shared with Mercurius in the form of a commission.
An investing product?
EV+ betting could be seen as a form of investment – a very risky one for sure, but also with no correlation to traditional asset classes. This could make it a suitable addition to an investment portfolio, under one important condition. Namely, that you manage to find EV+ bets, since anything else will destroy your capital in no time. Whether Mercurius Tradr can provide that certainly remains the most important question about the software.
Mercurius Tradr tries to identify value by analyzing and searching for inefficiencies in the 1X2 markets of big European soccer leagues. As already mentioned, these are some of the largest betting markets, so I certainly wouldn’t recommend trying this at home. Some surprisingly persistent trends notwithstanding, the 1X2 markets in the biggest football leagues on the planet remain a very tough market to beat. So what does Mercurius have against it?
Before diving into the technicalities, I think it is always important to have a look at who is behind such a project. The company was originally founded in Milan, Italy, by a young and diverse team with backgrounds in engineering, sports trading and design. Quite young, but also already with some experience in the field, the team was eager to prove that even the toughest markets can be beaten if you approach them with from the right angle.
Of course, in order to beat a large market, you require some serious resources, so the team went on to look for financing. They managed to do that getting an angel investor on board shortly after. That allowed them to buy a professional data feed and be able to set up and run their betting operation without worrying about the basics. Next thing, the team went in front of the drawing board to build their market-beating algorithm.
The Mercurius Model
The guys at Mercurius rely on a self-developed xG model, which tries to predict the outcomes of football games. The term xG model sounds kind of cool but remains a black box for most people who are not professional football analysts. So, even though I will try to present the technical part as well as I could, I will also use traditional methods to evaluate the picks that the model produces. CLV, p-values and the like are all concepts that punters have used since a long time to evaluate the quality of a bet and you don’t need to be a rocket scientist to understand them, so we will have a look at those as well.
But before that, here is a deeper look into the whole process and see how does Mercurius look for value on the big European football markets.
The input data
As mentioned above, the team managed to put their hands on a professional data feed for soccer, which would provide the input data for their model. The provider in question is WyScout (with some additional data from SportMonks). WyScout a company I have no experience with, but from what I know are one of the few premium services available, which offer a very detailed football data, way beyond what you would be able to find in most free football stats websites. Here you could be working with positional data, passes, possession, set pieces and interceptions as opposed to just basic stats as shots on target, goals or cards.
I must also say I have seen comments on Twitter by some industry insiders claiming that the feeds of Opta and StatsBomb are of a better quality in terms of verification and correctness of the provided numbers. However, they also seem to be much more expensive, so at the end this is the trade-off that you must make when you are just getting started.
In any case, I believe a professional data feed is a prerequisite for building a winning model in such big leagues, so it is definitely a very useful (and necessary) tool.
EDIT 18.3.2020 18:11 CET: The Mercurius team commented on this point, that StatsBomb even though having great data did not cover all the leagues they are interested in. Furthermore, Mercurius have strategic partnership with WyScout as well as short lines of communication (due to them being Italian company as well) so WyScout was the natural choice.
Lorenzo Malanga, Head of Trading at Mercurius, is the tech guy, who is overseeing the process of model development and applying his knowledge from the financial industry in order to squeeze some edge from the data. How does he do that?
The below video gives a very brief intro into the xG methodology and how Mercurius uses it:
The Mercurius model is basically an xG (expected goals) model. xG is a very trendy methodology, which is a hot topic in the football analytics and the betting communities alike. With the recent emergence of more detailed data it became possible to calculate the expected outcome of any event in a game by comparing it to similar events from previous games. Mostly you would be looking at shots and where those shots were made from. Then you would compare with your database to determine how often does a shot from that location turn into a goal. If, say, 15% of shots from this location are being converted, your team has just scored a 0.15 xG. At the end of the game you compare the xG of the two sides and see which one performed better.
Why would you need that?
Of course, you could alternatively just look at the end result. However, football is by design a sport in which chance plays a huge role. Being the most low-scoring of all big sports, luck is a major ingredient of the end result of the game. Therefore, you could argue, the end result of a game does not tell you that much about which team played better. xG reduces the element of chance by adding every scoring situation that occurred in a game for what it is worth in terms of scoring potential.
Convert to betting odds
xG could tell you which teams performed above or below their results in recent times, which could be a solid base to make more accurate odds for their next games. This is the methodology that Mercurius relies on.
It is a two-staged approach – you first quantify each specific action in a game and assign to them a xG-number or a sort of a rating. Than in the second step, you correlate all the events in the game to expected goals in future games. This includes estimating the strength of the teams, measuring time-decay effects and so on. You end up with a probability for each possible result, which adds up to the probability of the three different outcomes – home win, draw or away win. Those are being calculated using a variant of a Dixon & Coles model.
Once the value has been discovered, a bot places your bet automatically at Betfair. Obviously, as volume grows liquidity can turn into a problem. According to the Mercurius team, they don’t have an issue with that yet. However, to be able to accommodate future customer growth, the team is in discussions with broker firms in order to find ways to place larger volumes without disturbing the market. Specifically, the estimate of the Mercurius team is that the current market liquidity can handle €3-5M, while they are currently at around €700k.
Bet record / User interface
You get complete bet record of the bets the software has placed for you in your Mercurius account. It looks like this:
In the overview you can see where your ROI and ROC lie at this point.
What is my current performance?
I have joined the service in June 2019 depositing GBP 1000 real money to play the Mercurius tips and test the software. At the time of writing this article, the tool has made 470 on my behalf with a ROI (return on investment / turnover) of 0.51% and a ROC (return on capital) of 2.52%. This equates a current level of my bank of GBP 1025.17.
What do those results say? Does the software identify value? For sure with this performance and the number of bets the question cannot with certainty be answered affirmatively. However, and to be fair, that was never really a surprise for me.
Sample size issue
In truth, 500 bets are not nearly enough to evaluate the profitability of a service beyond a shadow of doubt. Going by a regular t-test, the current positive performance has 37.8% probability of happening by chance. Obviously, that is not nearly enough to discard the null hypothesis (which in betting is normally EV-). The evaluation is made harder by the fact that the software more often identifies value in long odds than in short ones, meaning you would need a yet larger sample in order to make a strong conclusion.
For example, again using a p-value to judge a theoretical larger sample of bets with the same ROI figure, you would need almost 14 000 bets to be above 95% confident that the performance is not a result of chance! To achieve that number you would need years using the service and it would obviously not be practical for me to wait for that long in order to write the review.
I knew that would be the case at the time of writing already when I joined the service. Therefore, I can say I currently don’t have the tools to evaluate the profitability of the service with complete certainty. The small profit so far is a positive indicator for sure, but it remains to be seen how it goes in the future.
What about CLV?
Regular readers would know that you can evaluate the profitability of a series of bets even with smaller sample like mine. You do this by measuring the CLV (closing line value). The CLV tells you how much on average have your odds drifted after backing the selection. Drifting odds of a backed selection are a good indicator that the market sees value there as well. Studies on the CLV have shown, that even within a small sample of several hundred bets you can make a very reasonable evaluation regarding the value contained in those odds.
Unfortunately, I don’t have access to closing prices on the advised bets, so I can’t calculate the CLV of the series. Perhaps this would be something that the Mercurius team can implement in the future. I believe CLV is not only a very informative and useful metric for this type of betting strategy, but also a good CLV will help attract much more new customers than any form of marketing ever could. Perhaps a point to consider.
The staking that the strategy is following is a 1% of bank stake for every bet. The team prefers not to apply a Kelly staking plan, as they consider it too risky for most people. Therefore, they decided level staking is a better option here. While I agree that Kelly might be too risky for most people, I think partial Kelly would still be acceptable option for most (in one of my current betting models I am using 1/8th Kelly).
Then again there is the valid argument that perhaps with the chosen model you cannot be sure about the size of the edge of your bets, so you lack the input data for properly calculating a Kelly stake. However, if this is the case, fixed profit is a better staking strategy than fixed stake, as it basically represents Kelly staking for unknown edge and maximizes bank growth.
At the end of the day it is a decision of the Mercurius team what they prefer, since they make their fees from their customers performance (will talk about it a bit later). I still believe an optimal staking plan would be beneficial for everyone.
Furthermore, allowing some liberty to the customer to adjust the stakes would be a beneficial feature. For me, since the capital staked per game is a relatively small proportion of my bank, at the end I did not really need the whole amount I have deposited. I could have made the same bets with the same stakes, having locked in only half the capital. Of course, the profit/loss would look differently in relative terms, but I can live with that, and perhaps it would have been better if I had the opportunity to go this way.
EDIT 16.3.2020 21:24 CET: I received feedback from Carlo from Mercurius, that you can set your capital to a certain number and still withdrawas much as you see fit, so the above really shouldn’t be an issue.
My experience with the tool
The tool really works as seamlessly as advertised. I had zero need to login after making the initial deposit. Without having intention to withdrawal, there is really nothing for me to do there. And given the volatile nature of betting, it is never a good idea to obsess over the current size of your bank. The best strategy usually is to sit back and wait till you have a reasonably large sample to make assessments. The software allows you to do just that.
The pricing of the product was a part of it which I really liked. The cost structure is like that of a financial fund, whereby you mostly pay for the overperformance of your investment. This is quite innovative for the betting markets and something I personally haven’t seen so far elsewhere. Paying for performance is most definitely the fairest form of a fee, so I can just congratulate the Mercurius team for going this route.
There is a fixed component as well, but after 6 months of using the service I have found it to be quite negligible. After receiving my first bill for using the service, my annual fee was at 2% of my initial deposit, which is exactly 20 GBP. One could live with that.
The performance fee on the other hand is 30%, however only above a hurdle rate of 5% ROC (return on capital). In other words, before reaching 5% total return of your investment, you don’t pay any performance fee. Additionally, a high watermark is applied, meaning commissions are only deducted after the previously highest bank level has been achieved. This is also typical for financial funds and ensures that all losses are being recouped before charging the customer for overperformance.
Way better than the market standard
Most tipsters out there would charge you a fixed fee for a period, without regard for the number of tips made or the money you play them with. This leads to tipsters making fewer and fewer tips each period, effectively increasing your cost per pick. Even worse, if you play with smaller stake than other customers of the service, the whales often bring the odds down significantly before you are even able to place your bet. All that while you pay a significantly higher fee per dollar bet than them. This is how most tipsters roll and is indeed a cost structure with many disadvantages for the punter and zero redeeming qualities.
Then recently a new type of services arrived, which charge you based on the volume you bet. Smartbet works on this principle, as I suppose some or most of the services, which like Mercurius are partnering with Betfair. There the whales are paying a high fee and the small stake players a low one, according to the bet size. This is already a lot fairer than a fixed period fee and was already a great innovation on the betting scene as far as tipster pricing is concerned.
Skin in the game
The performance-based fee model of Mercurius beats even that. Here you do not just pay based on your volume, but your fee is adjusted yet again according to the investment performance. Choosing this fee structure is very rare, since the income of the service depends on the performance of its picks. The fact that Mercurius chose it shows that they have a lot of confidence in their concept and are “in it” together with their customers. As N. Taleb would put it, Mercurius does have skin in the game. Overall, this is the fairest cost structure I have ever seen in the betting space and I can congratulate the guys over at Mercurius for choosing to go with it.
The very reason I have decided to give Mercurius a try is that I like their concept, including the approach to finding value, the courage to attack a large betting market with a proprietary AI model and the innovative and fair fee structure. I continue to see value in the business model and will be excited to follow the development of the algorithm and Mercurius as a company in general.
The sample I have collected shows perhaps some indications that the model finds value on the market, though no conclusions, positive or negative, can be drawn from a sample with only ~500 bets. For me the long-term process of finding value is more important and it looks promising. One of the prominent US punters Rufus Peabody has named his podcast, “Bet the Process” and I do believe this is the proper approach to any betting strategy.
It is perfectly understandable that a potential customer would like to have more certainty about the current profitability of the model. There is a way to show that in a small sample – you just need to calculate a CLV. No other marketing tool can attract as much betting capital as proving a positive CLV. It is certainly a strong recommendation I can make to the Mercurius team. Alternatively, a long winning sample might do the same work, but you need quite a bit of time to achieve that.
That aside, the Mercurius betting model looks very promising. The strategy is easy to setup. The team certainly does not lack the skills, motivation or resources needed to continue to further improve their product. The pricing is very reasonable and fair. If long-term value betting is your thing, I would certainly recommend getting familiar with the Mercurius concept and perhaps even giving it a try. This is what I did, and I was left with a very positive impression.
These were my two cents on the Mercurius Tradr tool. I hope you enjoyed the review. Feel free to drop any questions you might have about the product in the comments below. I will do my best to answer them and I am sure the Mercurius team will also monitor the post and give feedback if needed.
On an unrelated note, it seems that just about all sports event were cancelled due to the COVID-19 outbreak, so there probably won’t be much to bet on in the following weeks, maybe even months.
That is bad news for my bank but does not mean that I cannot write about betting. I intend to use the time freed up from betting activities to do just that.
Hopefully that will provide a little distraction from the predominantly grim news you read just about anywhere these days. It seems that a killer virus pandemic can be a bit of a downer. As Jürgen Klopp, I am not an expert on viruses so I can’t say much on the topic. Perhaps just that: stay self and sane, apply some social distancing, avoid crowded spaces and think about the most vulnerable ones. If we all remain reasonable and take care of each other, I do believe we will soon have all this behind us. So, keep the spirit high, and till the next time!