Backing Tennis Favourites vs Underdogs: What 3,273 Real Bets Revealed

Tennis appears to offer one of the simplest betting decisions in sport. Unlike football, there is no draw to consider. Two players enter the match, one wins, and the other loses. The betting market identifies a favourite and an underdog, with prices reflecting their perceived chances of victory. Surely, backing the stronger player consistently should produce a reasonable outcome? And if the favourite is too short, perhaps the outsider offers better value. During October 2026, the Trade Carefully Research Lab examined both propositions using thousands of actual bets. What emerged was considerably more interesting than a simple contest between favourites and outsiders.

Our tennis experiments recorded 3,273 settled bets across ATP, WTA, Challenger and ITF competitions. These were real bets placed in live exchange markets using actual money, not simulated backtests, historical price reconstructions or hypothetical selections. The stakes were deliberately small because our objective was to collect reliable evidence about how different approaches behave in real market conditions. We were testing ideas, not trying to generate an income from unproven strategies.

Two experiments ran alongside each other. One backed the market favourite, while the other backed the outsider. Both operated systematically across a broad selection of tennis competitions rather than relying on discretionary match predictions. By examining the results together, we could investigate whether either side of the market appeared to offer better value.

The headline figures were revealing. The favourite experiment recorded 1,637 bets and 1,047 winners, achieving a strike rate of 63.96%. Despite winning almost two-thirds of its selections, it lost 53.21 units after commission, a negative return of 3.25%.

The underdog experiment recorded 1,636 bets and 571 winners, a much lower strike rate of 34.90%. Yet it finished almost exactly level, making just 0.23 units after commission, equivalent to a return of approximately 0.01%.

The contrast is striking. The favourite experiment won 476 more bets but produced a substantially worse financial result. The underdog experiment lost nearly two-thirds of its selections yet avoided an overall loss. This illustrates one of the most important principles in betting research: the number of winners tells us surprisingly little without considering the prices paid for them.

A favourite backed at 1.30 produces just 0.294 units of profit after 2% commission if it wins, but loses a full unit if it fails. Several successful selections can therefore be wiped out by a single defeat. The underdog has the opposite problem. It loses more frequently, but a winner at 5.00 produces 3.92 units after commission, enough to compensate for almost four unsuccessful selections.

Neither side is inherently better value. Everything depends on whether the prices accurately reflect the true probability of winning. That is what our research is trying to understand.

ATP Tennis: An Unexpected Advantage for the Outsiders

When we separated the results by competition type, the ATP findings immediately stood out.

Across 316 ATP favourite selections, the experiment recorded 193 winners but lost 24.79 units, representing a negative return of 7.85%. The ATP underdog experiment, meanwhile, recorded 119 winners from 312 bets and finished 40.04 units ahead, an impressive return of 12.83%.

That is a remarkable difference between the two sides of the same broad market.

The ATP Shanghai tournament provided some particularly interesting observations. Our underdog experiment recorded a winning selection in the match between Walton and Te at odds of 16.50. Another recorded winner came from the match between Zizou Bergs and Carlos Alcaraz at odds of 12.00. These two winning bets alone contributed approximately 25.97 units after commission.

Shanghai was also one of the most profitable tournaments for our underdog experiment, generating 34.68 units from 77 selections. Meanwhile, the favourite experiment lost 11.51 units across 68 selections at the same tournament.

These results are fascinating, but they also demonstrate why caution is essential. A couple of unusually high-priced winners can transform a tournament's overall return. Removing them from the analysis would be misleading because they were genuine winning bets. But ignoring their influence would be equally misleading.

We cannot yet conclude that ATP outsiders are systematically undervalued. We can say that during October, our recorded ATP underdog bets substantially outperformed the favourites, and that the result deserves further investigation.

WTA Tennis: The Pattern Changes

The women's tour produced a different picture.

Our WTA favourite experiment recorded 340 winners from 522 selections and lost 5.40 units, a relatively modest negative return of 1.04%. The WTA underdog experiment recorded 175 winners from 532 selections and lost 15.62 units, equivalent to a negative return of 2.94%.

Unlike the ATP sample, backing the outsider did not produce a positive overall result.

Individual tournaments nevertheless showed considerable variation. At Wuhan, the underdog experiment returned 30.77 units from 92 selections, while the favourite experiment lost 8.64 units from 89. At Osaka, the situation reversed dramatically: favourites gained 12.79 units from 50 selections, while outsiders lost 24.49 units from 52.

The Osaka underdog loss represented a negative return of approximately 47.09%, an extremely poor result for a single tournament. Wuhan, by contrast, produced a positive underdog return of approximately 33.45%.

How can two tournaments on the same tour produce such different outcomes? The immediate answer is that tennis results are variable, and relatively small tournament samples can be heavily influenced by individual upsets. Other possible factors include the strength of the competing fields, player fitness, match conditions and the prices available. Those explanations remain hypotheses, not findings established by this dataset.

The important point is that the overall WTA figure conceals substantial variation. A broad result tells us what happened in aggregate, but it does not explain why.

The Odds Tell an Even More Interesting Story

Perhaps the most revealing part of the tennis research emerged when we examined underdogs by their matched prices.

The underdog experiment recorded 935 selections at odds between 2.00 and 3.00. These produced 408 winners and a positive return of 22.16 units, equivalent to 2.37%.

Between 3.00 and 4.00, the experiment lost 4.75 units across 320 selections. Between 4.00 and 6.00, the loss increased to 18.47 units across 198 selections.

Then something unexpected happened.

The 98 selections matched between 6.00 and 10.00 produced 20 winners and finished 44.70 units ahead, a return of 45.61%. That is an extraordinary result for an odds range in which winners are relatively uncommon.

But moving beyond odds of 10.00, the picture deteriorated sharply. Across 73 selections at prices of 10.00 or greater, only two won. The resulting loss was 45.03 units, a negative return of 61.68%.

In other words, the strong performance between 6.00 and 10.00 was almost completely offset by the losses on even longer-priced outsiders.

This is precisely the sort of pattern that makes systematic research worthwhile. Looking only at the overall underdog result of +0.23 units would reveal almost nothing. Examining the distribution shows that different price ranges experienced dramatically different outcomes.

But there is a trap here. We could look at the profitable 6.00–10.00 range, discard everything else and announce that we had discovered a winning tennis strategy. That would be retrospective selection: choosing the best-performing part of a dataset after already knowing the results.

A genuine advantage must survive fresh observations. If the same odds range continues to perform well over subsequent months, we can begin investigating whether there is a persistent pricing inefficiency. Until then, the result is an interesting signal, not a proven edge.

Challenger Tennis: The Less Glamorous Reality

Away from the major tours, the Challenger competitions provided another useful perspective.

The favourite experiment recorded 676 selections and lost 23.30 units, a negative return of 3.45%. The underdog experiment recorded 679 selections and lost 9.43 units, a smaller negative return of 1.39%.

Neither approach was profitable overall.

The individual tournament results varied considerably. At the Fort Worth Challenger, favourites lost 8.46 units across 42 selections while underdogs gained 3.40 units across 44. At the Florence Challenger, favourites lost 8.96 units across 38 selections. The underdog experiment at the Guangzhou Challenger lost 12.71 units across 39 selections.

These results reinforce the need to distinguish between the broad category of tennis and the individual markets operating within it. An ATP Masters event, a WTA tournament and a Challenger competition may all offer a two-outcome match market, but that does not make their pricing behaviour identical.

The evidence is not yet sufficient to establish which differences matter most. What it does show is that treating every tennis match as an interchangeable betting opportunity can conceal substantial variation.

Why We Use Real Bets Rather Than Backtests

There is an important methodological distinction behind the Trade Carefully Research Lab.

Many betting strategies are presented using historical simulations. A researcher takes past results, applies a set of rules and calculates what might have happened. Such backtesting can be useful for exploring ideas, but it has limitations. Historical prices may not have been available at the moment a bet would actually have been placed. Market liquidity, execution timing and commission can also change the outcome.

Our approach is different.

We place actual small-stakes bets in live exchange markets and record the matched prices and settled outcomes. 

We are therefore observing what happened when the experiments were genuinely executed, rather than calculating returns from prices we might theoretically have obtained.

That does not make our research immune to bias or guarantee that the results will continue. Small-stakes execution may not scale to larger amounts, and a few months of observations cannot establish a permanent market advantage. But it does remove an important source of uncertainty: whether the recorded bets were actually achievable.

The small stakes are intentional. We want to gather evidence across numerous ideas without taking substantial financial risks on approaches that have not demonstrated a reliable advantage. A collector that loses money can still be valuable if it teaches us something about market behaviour.

In fact, the favourite experiment's loss of 53.21 units is arguably one of the most useful findings in this study. It challenges the intuitive assumption that winning nearly two-thirds of tennis bets should be sufficient to generate a profit.

The market does not reward accuracy alone. It rewards accuracy relative to price.

What Have We Actually Learned?

October's tennis results provide several reasons to question simple betting assumptions.

Backing favourites won considerably more often than backing outsiders, but produced a worse financial result. ATP underdogs performed strongly, while WTA underdogs lost money. Challenger markets delivered losses on both sides. Moderate outsiders between 2.00 and 3.00 produced a small positive return, while those between 6.00 and 10.00 generated an unusually strong result that was almost entirely offset by the longest-priced selections.

None of these observations proves a dependable betting advantage. They do, however, demonstrate why blanket approaches are unlikely to tell us enough about where value might exist.

Our growing suspicion is that blindly backing selections across every tournament, price range and market condition is unlikely to provide a sustainable advantage. The October evidence supports taking that concern seriously, although it does not establish a universal rule. Some broad experiments finished positive, others negative, and the differences between subsets were substantial.

The temptation is to search for a quick winning formula. Our objective is different: identify what consistently fails, understand where unusual results occur and test whether any promising patterns survive new data.

The wider Research Lab began collecting observations around 23 September 2026. Once those earlier records are incorporated, we will examine the cumulative evidence rather than relying solely on October. Every subsequent month will add another opportunity to challenge the apparent patterns.

If ATP outsiders continue to outperform, that will be worth investigating. If the profitable odds ranges reverse sharply, that will be equally informative. We have to be willing to accept either outcome.

There is no shortage of persuasive theories in sports betting. What is much harder to find is a transparent record of what happened when those theories were tested with real money, including the losses.

That is the purpose of the Trade Carefully Research Lab.

The data does not tell us what we want to hear. It tells us what happened.

And sometimes, the most important finding is discovering that being right most of the time still isn't enough.

Research note: This analysis covers 3,273 settled tennis bets recorded during October 2026. All experiments involved actual small-stakes bets placed in live exchange markets. Results are standardised to level one-unit stakes using recorded matched prices, with 2% commission deducted from winning profits. Competition and price-range findings are exploratory, not evidence of established profitable strategies. Past performance does not guarantee future results. Betting involves financial risk.