Interactive Forum December 2025: Joao Fonseca: Forehand

[giancarlo]

Brazilian Joao Fonseca is only 19 years old and already turning heads with one of the most compelling forehands on tour. According to the ATP’s database, “Beyond the Numbers,” Fonseca’s average forehand speed is 81 mph (130 km/h), compared to the tour average of about 76 mph (122 km/h), and his spin rate is 3,019 rpm, significantly above the tour mean of approximately 2,844 rpm.

In October, Fonseca won an ATP 500 event in Basel, Switzerland, beating 18th-ranked Alejandro Davidovich Fokina, 6-3, 6-4. This was a major milestone in Fonseca’s rise – his first ATP 500 title (and the second of his career). Fonseca’s ascended this year like a supernova – from a ranking of 145 in the world at the end of ’24 to a career-high ranking of 24 early November.

The Fonseca forehand is built on modern mechanics:

• large unit turn, aided by his left arm staying on the racket for -- a complete turn, reminiscent of Carlos Alcaraz

• strong leg drive

• making use of all stances

• fast racquet head acceleration, a follow-through release that generates penetrative ball flight.

Fonseca’s blend of pace and spin pushes his forehand into the same conversation as the weapons of Carlos Alcaraz and Jannik Sinner.

Fonseca’s ball quality — not just how hard he hits it, but how effectively he uses it — is already at elite levels. Once ’26 gets underway, we await his progress on point construction, mental management of big moments, and how he fares in week-in and week-out competition.

Your thoughts please.

[stotty]

Yet another bazooka forehand hits the tour. The place is awash with them. This one is that bit better than most perhaps because he gets so much spin and incredible pace...usually pace is sacrificed to spin. I love the wide base, deep turn, and the way he drives up into the ball. He has excellent footwork and clearing steps too. I think we are all hoping Mr Fonseca is going improve enough to eventually challenge Sinner and Alcaraz. Fingers crossed.

[stroke]

https://www.atptour.com/en/stats/tdi-leaderboard?timeFrame=average52Week&versusRank=all&surface=all&sortBy=ShotQuality.Forehand&sortOrder=dsc

Per tennisinsights, Fonseca forehand is number 19, but he definitely looks like a candidate to move up

[jimlosaltos]

https://www.atptour.com/en/stats/tdi...&sortOrder=dsc

Per tennisinsights, Fonseca forehand is number 19, but he definitely looks like a candidate to move up

While I prefer actually measuring things my own eyeball test says Joao's forehand is well above 19th in world.

Then, perhaps arguing against myself, I remember last year's NexGen where Joao meet Learner Tien and the young SoCal guy, was handling Joao's cross court forehand rallies to his lefty backhand with ease. Then, Joao changed tactics and started blasting up the middle, creating space and took over.

Forgive me a digression, but the ATP having two sets of stroke rankings seems silly but the differences between them are perhaps revealing.

For example Jannik Sinner's serve ranks number one in the ATP stats based on results {the little ranking box on the ATP front page, right}, but the TDI stats (based as I understand it on measuring the shots on speed, spin, placement, elevation etc) puts SInner's service at only 24 in the world, behind the likes of Korda and Griekspoor. To mix senses, that doesn't pass the smell test.

This a long-winded way (sorry) of getting to my point that the TDI says that a forehand going higher over the net is "better" than one with little clearance. Both Joao and Sinner hit heavy spin but without the high trajectories we expect from Rafa's school of forehands. They don't need a big arc over the net to be reliable, instead aiming to penetrate the court. So, is TDI penalizing their forehand ranking because they focus on taking time away from opponents by hitting a lower ball? That's my guess.

[stroke]

Jim, Sinner does have the number 1 forehand per tennisinsights. I personally think his ability to take the ball on consistently so early is in the mix. I would have loved to have seen the tennisinsights forehand rankings when Fed and Nadal were doing their thing. My guess is Sinner is probably the first player ever to be the best off both wings. Maybe Agassi around the early 2000's.

[stotty]

https://www.atptour.com/en/stats/tdi...&sortOrder=dsc

Per tennisinsights, Fonseca forehand is number 19, but he definitely looks like a candidate to move up

Fonseca is still only 19 years old so plenty of time to move the ladder with that forehand. As I've said before, stats can reveal a lot but they can also be misleading.

[stroke]

Fonseca is still only 19 years old so plenty of time to move the ladder with that forehand. As I've said before, stats can reveal a lot but they can also be misleading.

I certainly agree with that Stotty, but I still trust the tennisinsights stats/rankings more than my own eye. After the top 4, the Performance Ranking numbers have a bit of a drop and are all pretty lumped together. I certainly think Fonseca could end up in the top 4.

[jimlosaltos]

Fonseca is still only 19 years old so plenty of time to move the ladder with that forehand. As I've said before, stats can reveal a lot but they can also be misleading.

Agreed. Let me take another try at explaining my point. We don't know TDI's recipe or how it compiles these stats to get its ranking, just the ingredients. IF, as their words imply, they give a higher ranking to a forehand that has greater "clearance" over the net, they may be falsely penalizing Sinner's forehand (and perhaps Fonseca's) for something that is actually a strength. i.e. That aggressive often (surely not always) lower trajectory WITHOUT lots of UFEs is a comparative advantage.

[stroke]

https://www.instagram.com/p/CppKuJEo4KS/?igsh=MWt2czdiZHd6anpzbg==

[stroke]

https://www.instagram.com/reel/C4isJrsu-vI/?igsh=MXRkMzc3c3R6NTV6bg==

This video shows shot quality numbers on an extended rally Sinner vs Alcaraz

[stotty]

https://www.instagram.com/reel/C4isJ...c3c3R6NTV6bg==

This video shows shot quality numbers on an extended rally Sinner vs Alcaraz

Interesting to get some insight how shot quality is measure. So length and pace and direction are clearlt the biggest factors? I didn't see a good defensive shot in that rally but I wonder how a good defensive shot would score...slow but with height and depth. Not sure I would have given Alcaraz's drop shot 5...more like 3...not his best one.

[stroke]

Interesting to get some insight how shot quality is measure. So length and pace and direction are clearlt the biggest factors? I didn't see a good defensive shot in that rally but I wonder how a good defensive shot would score...slow but with height and depth. Not sure I would have given Alcaraz's drop shot 5...more like 3...not his best one.

The How Shot Quality is Measured link I posted to me indicates that there are so many factors going into it, it is impossible to say what the biggest factors are.

[jimlosaltos]

The How Shot Quality is Measured link I posted to me indicates that there are so many factors going into it, it is impossible to say what the biggest factors are.

This likely gets outside the bounds of what a TPN forum reader cares about, but if TDI is genuinely using AI to evaluate shots as claimed, it could well be a "black box" where they don't actually know what what the weightings are. Neural Networks learn answers from observing data to discover patterns they later recognize. Figuring out what those self-created patterns look for after the fact is difficult.

So, the TDI metric "could" be punishing Sinner and Fonseca for patterns of play that work less often for others, but exceedingly well for them { say, forehand clearance}. Maybe. Regardless, I have no idea why one set of stats based on actual performance (the ATP front page) rates Sinner as the best server while the TDI stats put him way down at an illogical ( to me) 24th behind the likes of Korda and Griekspoor {I think I switched forehand for serve on an earlier post, if so, apologies.}

I'll try a crude example, within my limited understanding, of how this lack of ability to "see inside the neural net the black box" can be a problem and hope I'm not boring people to tears. Building databases of X-Rays to train a neural net on medical diagnosis poses challenges that are less problematic than, say, autonomous driving of cars, where enormous databases are being collected. There are relatively few X-rays available {digitized, HIPPA, propriatary etc}. With a smaller sample, iff an AI says "I see a pattern, that means this is a correctly diagnosed tumor" is it doing that because of the shades in the image, or because 1) It was diagnosed on a Monday when there are fewer erroneous diagnoses than on Friday (true phenomenon) or 2) Because it recognizes the doctor's name and has figured out his diagnoses are often right? Nothing comes out of the system to explain that. Huge amounts of data would tend to reduce that type of error.

So, AI models have filters called Reinforcement Learning from Human Feedback (RLHF) to correct known problems (recognizing skin color as human) or deliberately skew results for political reasons, like what Musk does on X aka Twitter. Too long. Sorry.

Back to tennis

[stroke]

This likely gets outside the bounds of what a TPN forum reader cares about, but if TDI is genuinely using AI to evaluate shots as claimed, it could well be a "black box" where they don't actually know what what the weightings are. Neural Networks learn answers from observing data to discover patterns they later recognize. Figuring out what those self-created patterns look for after the fact is difficult.

So, the TDI metric "could" be punishing Sinner and Fonseca for patterns of play that work less often for others, but exceedingly well for them { say, forehand clearance}. Maybe. Regardless, I have no idea why one set of stats based on actual performance (the ATP front page) rates Sinner as the best server while the TDI stats put him way down at an illogical ( to me) 24th behind the likes of Korda and Griekspoor {I think I switched forehand for serve on an earlier post, if so, apologies.}

I'll try a crude example, within my limited understanding, of how this lack of ability to "see inside the neural net the black box" can be a problem and hope I'm not boring people to tears. Building databases of X-Rays to train a neural net on medical diagnosis poses challenges that are less problematic than, say, autonomous driving of cars, where enormous databases are being collected. There are relatively few X-rays available {digitized, HIPPA, propriatary etc}. With a smaller sample, iff an AI says "I see a pattern, that means this is a correctly diagnosed tumor" is it doing that because of the shades in the image, or because 1) It was diagnosed on a Monday when there are fewer erroneous diagnoses than on Friday (true phenomenon) or 2) Because it recognizes the doctor's name and has figured out his diagnoses are often right? Nothing comes out of the system to explain that. Huge amounts of data would tend to reduce that type of error.

So, AI models have filters called Reinforcement Learning from Human Feedback (RLHF) to correct known problems (recognizing skin color as human) or deliberately skew results for political reasons, like what Musk does on X aka Twitter. Too long. Sorry.

Back to tennis

Certainly over my head. I did feel like I somewhat had a feel for the tennisinsights explanation of what they are trying to do.

[jimlosaltos]

Certainly over my head. I did feel like I somewhat had a feel for the tennisinsights explanation of what they are trying to do.

I think that's generally how people feel about AI, "I have a feel for what it's doing but I'm not entirely sure."

On the off chance the Venn diagram of TPN readers includes tennis and AI two links.

"Introduction to Generative Adversarial Networks", layman's terms, they're trained by putting two systems in sort of competition. One "hides" an image, the other "Finds" it.

And "Neural Networks (machine learning)"