#### The new tennis season is underway and the first major of the year starts in just a matter of weeks. They say that *‘three is the magic number’* so, in this post, I’m going to cover my method for finding and combining three solid selections into a single multiple bet. Let’s go!

Greetings and welcome back to another **tennis-related post**. If you have followed my blog for a while, you’ll know that I love to make money from tennis, be that **straight betting with bookmakers** or from **in-play trading on the exchanges**. You can find some previous posts on this subject here:

**How to Correctly Predict the Price Moves in Tennis Odds****Building a Tennis Betting Model****The Wimbledon Project****The Wimbledon Project – The Results**

But, the topic of today’s post is **multiple bets** and the methods by which I go about looking to combine some high probability *(but low odds)* selections into a more **attractively priced combination bet**.

Now, before we begin, I must stress that multiple bets of any kind are **highly speculative** for obvious reasons. As a result, betting in this way is **not a core strategy** of mine and not one I dedicate a particularly large bankroll towards. That being said, it can still offer** potentially very attractive returns **and a nice little side hustle that requires **little time or effort** with opportunities daily. I’ll show you how just a few low odd selections can be combined into a single bet, usually at odds well over evens.

**Which Tennis Exactly?**

What I am about to walk you through I personally only apply to **women’s singles matches on the main tour – the WTA** *(I ignore lower-tiered tours such as the ITF)*. You might also wish to try this with men’s matches on the main ATP tour. I’m not suggesting it won’t work, merely that I have not personally deployed it into the men’s game and I’m not one to advocate something I don’t personally do myself.

I find personally find that there are more than enough selections generated in the WTA and this is the main area of tennis I focus on due to various **statistical traits** of the women’s game compared to the men’s.

**Resources**

With that out of the way, what **resources do we need?** Thankfully, there is an abundance of good quality statistics available for free online. For this particular strategy, the key resource I use is the excellent site over at **tennisabstract.com**

Tennis abstract is a **great statistical resource** with a substantial amount of data that can inform a multitude of tennis betting or trading strategies. At its core, it is a database of player data capturing not only detailed results across the current *(and past) *seasons but also all manner of aggregated stats from player serve speeds; breakpoints saved or converted percentages to how often a player utilises slice shots! If you love stats as I do, you’ll want to use this resource, amongst others, in all your tennis betting and/or trading activities.

In addition to **tennis abstract**, we also need a quick and handy way to establish upcoming matches from which we can find suitable selections. For this, I use the always excellent **FlashScore**.

Here, I simply click on the **tennis menu** and you’ll see every match being played on a given day or coming days. In my case, I then look for **WTA singles matches** being played as my starting point for filtering out possible selections.

By way of example, as I sit here writing this post on a Thursday evening, FlashScores shows me there are **nine matches scheduled** for tomorrow’s WTA action across three tournaments currently being played this week. These all represent the latter stages of the WTA tournaments in **Auckland**, **Brisbane**, and **Shenzhen** – all key warmup tournaments for the upcoming **Australian Open.**

**Process**

The overall aim here is to find **high probability selections to combine into a multiple bet**. Rather than push my luck and go for crazy 10-fold accumulators bets, I tend to limit these to **3 solid selections** *(sometimes 4 at a push)* – three is the magic number after all 😉

So, I’ll focus on picking the very best 3 selections in my opinion and leave the rest. Let’s walk through this in more detail.

The first step is to use the excellent player data freely available on **tennis abstract**. I want a quick and easy way to appraise the likelihood of the **pre-match favourite** *(the player with shortest starting odds) *winning the given match-up. I want to judge this based on the **specific surface** the match is to be played and it’s also worth judging form and results in the context of **each player’s WTA ranking** and their respective history playing against **higher or lower-ranked opponents**. Tennis abstract allows us to do all of the above in a few simple clicks.

While **shocks can and do happen**, over a large sample size, **pre-match favourites win the majority of the time** – in other words, the bookmakers do a pretty good job at picking the likely winner in tennis matches. History has a habit of repeating itself and so we don’t want to go against that trend in the long-term. Looking through the lens of these stats can help us to filter out the very **best possible combinations and avoid those that look more like a coin-flip**.

## A Worked Example

So, let’s take the very first match shown on the schedule above – **Serena Williams versus Laura Siegemund** in the quarter-finals at **WTA Auckland**.

Clicking on this selection in **FlashScores**, we can see confirmation that this is a **hard court tournament** and that **Williams has a current WTA ranking of 10**, versus a **WTA rank of 73 for Siegemund**. We can also see *(unsurprisingly)* that Williams is the heavy **pre-match favourite** with a price to back of **1.10 with Bet 365**. Siegemund is the underdog here and priced at **7.0.** This translates to an **implied probability of 90.9%** **for a Williams win!**

This is a great example of a bet that you would likely **never want to take as a single**. At odds of 1.10 – the risk:reward is very unattractive – e.g., rising £10 to make £1. But, if we can use this in a combination bet it can help get us to an overall more attractively priced multiple bet.

While this is perhaps a **slam-dunk example**, and you may easily assume that Williams wins this match all day long without having to look at any stats, let’s at least confirm this with a glance.

Starting with Williams, I simply enter her **name into the tennis abstract search box**. You’ll see a useful summary page for this player and a wealth of useful information. Under the **recent results section**, I then click **‘all results’** to get a more granular view of her form.

You’ll now see a more comprehensive table of results for this player which can then be filtered further using any **combination of the drop-down filters** on the left of the screen.

Now, considering that this matchup is being played on a **hard court surface**, it would be more meaningful to see how Williams has fared on **hard courts**. A higher overall ranked player does not mean they are a given for a win on all surfaces. All players have their favourite playing surfaces and sometimes the contrast in results from one to another can be stark. So, viewing results specific to the surface being played on is, therefore, a **valuable piece of data** that the ranking alone does not account for.

So, the first filter I apply, using the drop-down menus on tennis abstract is to **filter the surface by hard court**.

I now see the record on hard courts specifically for Williams. Great. Let’s now consider that record in the context of matches against **similarly ranked opponents** to Siegemund. **Siegemund is ranked 73** in the world going into this match, so, using the **vs. rank filter**, I can filter the results for Williams to show her record against players ranked by one of the **pre-set choices**, or use my own. In this case, I’m happy to set it to players ranked **51+**. Using the pre-sets is great as we want to ensure the results are not too narrowly defined as that will result in a **very small sample size**. It’s also a time saver versus having to input specific ranges.

Lastly, I might also want to place greater weight on recent form compared to results from years prior. So, I can filter again by **time-span**. Most of the time, I go with the **last 52 weeks**. If the data sample size is too small, however, I’ll then consider the **total career for the time span**.

So, with those steps completed, I can see a better representation of Williams on this surface against certain ranked players. It shows me that Williams has a **9-0 win-to-loss record on hard courts against players ranked 51+ in the past year**. That’s a comprehensive record and exactly the kind of **dominance I like to see** when looking for statistically solid bets.

It suggests that, when facing a similarly ranked player on this specific surface, she stands a **very probable chance of winning the match comfortably**. Looking further at the breakdown of those matches, I can see that she has still lost a few sets along the way, so she certainly did not have it all her own way in recording these wins.

But, this is only half the story. Let’s now repeat the above steps, but this time for her opponent, **Siegemund**.

As before, I enter the player’s name into the search box, choosing to **view all results** and then applying the same filters; narrowing down results to **hard courts**, over the past **52 weeks** and, in this case, results against players **ranked in the top 10** – given that Williams is ranked 10th going into this match.

In doing so, however, this reveals that **Siegemund rarely gets to face a top 10 player** given her relatively low rank. Over the past 52 weeks, she **lost the only match she played on hard courts against a top 10 opponent**. In examples like this, I’m happy to widen the time span to **career so get a bigger sample size**.

Widening to see Siegemunds **whole career on hard courts against top 10 opponents**, we can see she has a **1-6 win-loss record** and that the single victory came from her opponent retiring. Skimming over these results we can also see that *(ignoring the set she won against an injured opponent which then retired)*, she has **not won a single set in all other matches** against top opposition.

So, considering **both player’s records** you’d likely wager that **Williams should win this match** based on past data. Of course, with the **starting odds of 1.10**, the bookmakers also feel this to be true.

Again, it may seem a trivial step, but there will be countless examples of matches where it’s **less clear-cut**, perhaps where the underdog has a history of taking big-name scalps, or where the favourite has a poor record on that specific surface, which her overall ranking may disguise. These are the potential red-herrings we want to avoid putting into a multiple. Sometimes success in betting is as much as what you don’t bet on than what you do.

I then simply repeat this analysis for each of the matches identified and group them broadly into **high, moderate or low probability** selections with a focus on combining the high or moderate ones into a multiple bet.

So, in completing the steps above for all **nine selections**, this is how I’d be ranking them for my combination bet. It is certainly not necessary to go to these lengths and write them out in this much detail, but I’m doing so to illustrate the process for you, the reader. Simply marking which matches to choose on a scrap of paper is sufficient 😉

So, in this example, I believe that combining **Williams**, **Kvitova**, and **Rybakina** into a multiple appears a solid choice. These appear to be the higher probability outcomes on this occasion.

An example of one selection that I’d avoid on my list would be **Julia Goerges vs Caroline Wozniacki**. Goerges is the favourite here but the players are very close in WTA rank and, looking at results on hard courts against players in that ranking bracket, **Goeges has a 16-10 win-loss record**. This is not bad by any means, but not exactly a home run. Equally, Wozniaki has a **13-6 win-loss record** against players in this ranking bracket, which would cast further doubt in Goerges to win.

Taken together, this would imply **a very close match that either player could win**. It is therefore hard to call with any confidence. The bookmakers also agree and have this priced at **1.80 for Goerges** and **2.00 for Wozniacki** – implied probabilities of 56% versus 50% respectively. These are the ones that might scupper a multiple bet and therefore one to avoid.

***** Update: Wozniacki won the match in straight sets, so we would have been right to avoid backing Goerges *****

**What To Look For And Avoid**

So, hopefully, it is quite obvious which combinations are attractive and which are not. I like to see obvious **signs of dominance for the favourite and weakness for the opponent** and then **avoid the less clear-cut matches** where the win-loss records are more evenly balanced. For example, trends such as a favourite with a 10-1 or 8-2 win-loss record is more appealing than 6-4 or 5-5 etc.

While it’s tempting to combine more than 3, I’ve found this to be the **optimal number.**

If that sounds like a complicated process, it’s really not. Once you have the hang of it, you can **quickly appraise any matchup in this manner in a matter of minutes** and move on. On a day like that shown above with 9 matches, this whole process might take **10 mins to quickly check over**.

Also, don’t feel like you have to combine matches on the same day either. I sometimes might find just 2 matches of interest on one day and find a 3rd on the following day to combine. I’m quite **picky with my selections** and focus on quality over quantity. **Be patient and don’t force it.**

## Combinations

So, let’s assume I go with the three matches that I marked as **High in my analysis** above. These three picks have starting odds with Bet 365 of **Williams** *(1.10)*; **Kvitova** *(1.44)* and **Rybakina** *(1.57)*.

While the bookmaker will **calculate the odds automatically**, to work it out yourself you can **simply multiply the odds together** *(in this case 1.10 x 1.44 x 1.57)* to arrive at your **combination odds of 2.49**. If I then placed a £25 bet and it was to win, I’d receive £25 x 2.49 *(£62.42)*, which, less the £25 stake, would give a **net profit of £37.42** or an **ROI of 149%.**

Of course, should that bet fail, we’d be sitting on a £25 loss 😉

So, that is my process in a nutshell for finding multiple selections. Once you have it down, it’s a **rinse and repeat exercise**. You’ll likely find many combinations on most days of the year, usually more in the early part of the week *(on account of more matches being played)*.

**Advanced Bonus Steps**

I’d argue that you’d have **moderate success just completing the steps above** to find half-decent selections to combine, but there are some **additional steps** one can take to add an extra confidence level, but with this comes some added time and complexity too. If that’s you, keeping reading.

**Beyond WTA and ATP Rankings**: Introducing Elo

While the **traditional ranking system** used in tennis serves as a respectable measure for the respective quality and form of a tennis player, many argue that they are an inferior method at predicting the likely winner of a match. This is largely because the points system which goes into the ranking number, does not take into account the quality of the opponent.

**Elo rankings** represent an alternative ranking system which is particularly well-suited to tennis. It is considered by many to be a **superior rating system** to the ranking formulas used by the ATP and WTA. The basic principle behind any Elo system is that each players’s rating is an **estimate of their strength**, and that each match *(or tournament)* allows us to update that estimate. If a player wins, her rating goes up; if she loses, it does down.

Where Elo really excels, however, is in determining the amount by which a rating should change up or down and this is influenced by the quality of the opponent. For example, if I was a top 20 player, beating the number 2 in the world should carry more significance than thumping a qualifier ranked 400 in the world. Elo rankings make those adjustments.

For a more detailed walk-through of Elo, check out this **excellent post**

**Elo 2.0 – Surface Adjusted Rankings **

Standard Elo rankings in tennis encapsulate **all matches on all surfaces**. While they are considered quite accurate in their own right, adjusting for the specific surfaces is widely regarded an even more superior measure and builds on the steps I introduced earlier in considering results on a specific surface.

Luckily **tennis abstract** also maintains detailed **Elo and surface-adjusted Elo rankings** for the top 250-300 players in both the men’s and women’s tours *(ATP and WTA)*. Factoring in surface-adjusted Elo rankings can provide a highly accurate measure for match analysis. I’m **certain that the bookmakers use surface-adjusted Elo rankings as a core input** when pricing matches.

So, using the highlighted matches from before, what do the **surface-adjusted Elo rankings** imply in terms of the likely winner? The degree of difference in the rankings of each player informs the probability of victory. For example, I may only want to make selections where the surface-adjusted Elo ranking gives me a **75% or greater implied probability. **

## Value Bets

So, we have the means to profile specific match-ups to consider the surface and WTA rank profile of certain opponents over a given time frame. We additionally have a **sophisticated surface-adjusted superior ranking metric** which gives a higher degree of probability of picking a winner. What other tricks do I have up my sleeve?

The surface-adjusted Elo rankings can be translated into their **implied probability** and, taken a step further, into **projected decimal odds** for both players. A final step you may consider is to measure those implied prices that the Elo rankings throw out against the advertised bookmaker odds *(i.e. looking for pricing differentials and potential value)*. If we only want to focus on picks with perceived value (a wise long-term strategy) we can use the data to establish which ones to combine.

For example, while I’ve no doubt **Williams** will win her match, the Elo surface-adjusted rankings imply she is **priced too short** at 1.10 on Bet 365 – the implied price based on Elo would be around 1.30. So, she would, therefore, represent be **a poor value bet** on this occasion.

Contrast this to **Keys. **We had already identified her as a moderate probability selection *(based on stats above)* and here we can see that the Elo surface-adjusted ranking differential is 201, which implies a **76% chance of her winning. **That translates into a projected price of 1.31 yet the price on Betfair 365 *(at the time of writing)* is 1.72. If we believe that a surface-adjusted Elo ranking carries any merit, this implies betting on Keys at an inflated price represents great value.

Of course, Elo rankings won’t be aware if Keys is **carrying an injury** so it’s always wise to check the news for any signs of concern about a player’s fitness. The Bookmakers are quick to re-price matches based on this sort of news.

So, if I focus on the matches with an **Elo surface-adjusted ranking differential of 75% or higher**; where the earlier stats support the pick and where there is **potential value in the price available with the bookmaker**, I’d be picking **Keys**, **Kvitova** and **Pliskova** as highlighted above.

**Automation**

I told you these bonus steps may appear complex. But, with relatively simple steps, a large part of this can be automated to save on time. **Google Sheets** offer numerous ways to directly import data from websites, such as with their **html import function** – a feature I use a lot to pull in data in the format I need.

For example, I automatically pull in the Elo rankings directly into my tennis spreadsheet. These update automatically each week and some simple formulas give me the implied probabilities for any match up together what Elo says the price should be. All that is left for me to do, is to enter the Bet 365 prices when they become available and my table pretty much highlights which matches to pick.

So, whichever method you choose, give it a try and let me know how you get on. Remember to keep stakes low as these are still multiple bets at the end of the day.

## Results

Oh, but before I go, you might be wondering how these picks turned out? Well, as I conclude this post, it’s now Friday morning and I can see the results on FlashScore as these matches were played overnight. And, of course, I also **put my money where my mouth is** and I placed the two multiple bets as outlined above via **Betfair Sportsbook** *(I’m restricted on Bet 365)* at similar odds. Let’s see how they did.

Starting with the first multiple, I placed a £20 multiple bet on **Williams** *(1.11)*; **Kvitova** *(1.44)* and **Rybakina** *(1.57)* at **combined odds of 2.52**. The match results are shown below:

Thankfully, you can see that these solid picks **all won in straight sets** and so my £20 bet returned me a nice £50.44, which when you take off the initial stake, equals a **net profit of £30.44** 🙂

Moving to the second multiple bet, using the more advanced surface-adjusted Elo rankings and value prices, I placed a **£25 multiple bet** on **Keys** *(1.70)*; **Kvitova** *(1.44)* and **Pliskova** *(1.41)* at **combined odds of 3.48**. The match results are shown below:

Again, this bet also landed and another set of straight-sets victories! For this one, my £25 bet returned me £86.97 for a net profit of **£61.97**. So, overall, that’s a net profit across both bets of **£92.41 for a total risk of £45**. Not bad for a few clicks of the mouse and a nice way to start my weekend. The completed bet slips for both selections is shown below.

Of course, **you won’t win them all** but I’d say you’ll win more often than not and with combined odds **typically well above evens**, it should yield profits in the long-run.

That’s it from me. Thanks for reading

Dan

## This Post Has 6 Comments

Thank you. Sounds a bit too easy for me at the first read as bookmakers have a lot of information to form their odds. But i will definately try it for a bit of time with some small stakes. Who knows. Maybe you found the golden formula 😛 Thank you anyways for sharing this with us.

Thanks Tim – I guess the point here is that we are siding with the bookmakers for that reason – i.e. they generally get it right most of the time because they have access and resources to know all there is to know. We are also backing and combining moderate to heavy favourites (typically) and which carry 75% or higher general probability of winning. I agree that the bookmaker price may differ from the Elo-adjusted implied odds due to superior information but it’s still a useful estimate of value in my opinion. Good luck

Given your reply to Tim above wouldn’t it be easier to check the bookies (researched and informed) odds and combine these into appropriate multiples rather than doing the research yourself?

I’m not trying to disparage your system which seems based on a sound rationale, just trying to genuinely understand the logic of putting all that effort in when, as you say, the bookies “know all there is to know” which is reflected in the odds they offer. Surely following their odds would therefore be allowing them to do the legwork when it comes to deciding on which bets to combine?

Thanks, CGC – If you believe in value betting (which many, myself included, would argue is the only way to bet) then simply taking the bookmaker odds at face value is a zero-sum game. We need to find value in prices, supported by data which suggests the bookmakers have it even slightly wrong, to win long-term.

After seeing your post I started looking at the elo ratings and trying to build a sheet to do the analysis, but I found it quite time consuming.

Looking at Tennis Abstract again today I realised I could avoid a large chunk of the work by just looking at the forecasts they run for each tournament, which are based on their elo ratings.

These forecasts are updated for each round, so I can just convert their published probabilities into odds and compare to the bookies odds to spot value.

As an example for the next round of the WTA Dubai they give number 1 seed Halep a 61% chance of winning her next match, whereas Bet365 have her at 2.1.

The advantage of doing it this way is it avoids the issue of your own personal interpretation of the stats clouding the issue. While this example might not be suitable for including in a multiple, you can still use it to spot value.

Hi Andy – thanks for your comment. I agree that doing the manual check is time-consuming, especially in the early rounds with so many matches. I have increasingly moved to a pure Elo-ranking approach. If there is a very close match, I might still check the stats out of curiosity. My sheet now is 90% automated, pulling in the surface-adjusted elo rankings across WTA and ATP, creating the matchups and then spitting out selections based on a couple of criteria I want to see in the rating differential – e.g. high surface-adjusted Elo win probability and where there is value to the bookmaker price (great for multiples); all value favourites; underdogs with value who have a fighters chance (e.g. 50:50 implied elo win probabilities) and also all elo underdogs with value. Really the only thing I have to do each day is input the bookmaker odds and I’m sure that too is ripe for automation. So, it’s essentially the same as the step you explain above.