Playing Tennis like "Deep Blue" Part 3: The Djokovic Effect

WARNING: This post is intended for Tennis Playing and Coaching Nerds. For all others, please move along, nothing to see here.

In the Deep Blue Part 2 blog on Tennis Computation & Tactics, we reframed tennis as a skills based tactical and strategic process rather than a psychological drama.

In the Tennis Scoreline Probabilities post, we turned that approach into a usable tool: a point‑by‑point model that lets players and coaches manually enter or randomly generate a set, visualize the Point Value of every point, and watch the leverage landscape unfold as peaks and valleys across the graph.

Those peaks aren’t emotional moments. They’re identifiable events that can be leveraged to win matches.

This post builds directly on that foundation and introduces the next step: using those Point Values to measure how well a player optimized their chances of winning.

This is a probabilistic approach to playing each point that we call GTO Tennis Play.

The Leverage Landscape: Tennis as a Value Map

If you’ve used the Tennis Scorelines Probability tool, you’ve already seen the leverage landscape in action. As you enter points, or generate a random set, the graph indicates the Point Value (PV) of each point from one point to the next.

Some points barely change your probability of winning. Others spike upward into steep peaks that can make all the difference.

Those high peaks are the "Big" points not because commentators say so, not because the crowd gets loud, but because the formal system underlying tennis makes them worth more.

For instance, winning or losing the point on your serve at 1–1, 0–0 will shift your win probability by 8% as shown below:

Whereas winning or losing the point on your serve at 5–5, 30–40 will shift your win probability by 39%.

We refer to these probability values as Point Values, and if you want to know more about how the Point Value model works, see Playing Tennis Like Deep Blue Part 2: Computation and Tennis Tactics.

The short explanation is that the Point Value model was produced using a Monte Carlo style simulation with the point, game and set structure, otherwise known as the formal system underlying the game of tennis. This simulation was then used to simulate millions of games between two players with equivalent skills, tactics and strategies.

The frequences of wins and losses that came out of these millions of simulated games were then used to calculate the probabilities for each of the possible scorelines.

The Tennis Scoreline Probabilities tool is therefore able to graph out the probability terrain for each and every set that a tennis player might end up navigating.

Once a player can see and understand this probability terrain, the playing of a tennis set becomes clearer, more structured, and the player suddenly has a better way to strategize around each point.

The Point Value Model: A Quick Refresher

The Point Value (PV) model calculates, for every point in a set, how much a player's win probability changes. Point Value is the difference between the player's win probability:

if the player wins a particular point
and if they lose that particular point

If you think about Point Value in terms of a poker game, it is the size of the pot that the player can win or lose with the playing of each point.

You can explore this in the Tennis Scoreline Probabilities tool by either:

manually entering each point for an entire set
or generating a simulated set where every point is randomly determined
and then analyzing the Point Value graph to understand the leverage structure of that particular set

You can see the PV for each point either by looking at the highlighted row in the table shown in the background of the Tennis Point Value tool, or by hovering over the PV point in the graph.

The peaks in the PV graph show you where your probability of winning the match can swing the most, while the valleys and plateaus show you where your probability of winning the match moves very little.

GTO Tennis Play: What It Means (and What It Doesn’t)

In poker, Game Theory Optimal (GTO) play means:

maximizing expected value
allocating your best decisions to the highest‑value situations
avoiding patterns that can be exploited
adjusting aggression based on the size of the poker pot

Although tennis does not have a solved equilibrium strategy, it does have the same underlying structure - a Point Value structure. Players who optimize their play on the Point Value of each point are effectively playing a GTO‑style strategy.

By using their tactical and strategic acumen, a player can then maximize the Point Value they accumulate from the points that matter the most.

Some players do this inconsistently. Some do it intuitively. Some, like Djokovic, appear to do it with conscious and practiced reliability.

And now we introduce a way of measuring and understanding how well a particular player has optimized Point Value during the playing of a match.

The Player’s Value Ledger

With each point you add to a set in the Tennis Scoreline Probabilities tool, the tool calculates both the PV for that point, and the sum of the player's wins and losses of PV up to that point.

consistently winning more PV increases the player's probability of winning the set
consistently losing more PV increases the opponent's probability of winning the set

This summation of won and lost PV throughout a set is a quantitative measure of how well the player is playing compared to their opponent.

It is essentially the same as tallying up the total amount of chips that a poker player has accumulated at the end of a poker playing session and is likewise a tennis analog to a poker session ledger.

Turning Value into Optimization

The total amount of Point Value in a tennis set varies over a wide range depending on the particular set of scorelines (individual points) that it includes.

The Point Value model contains all possible tennis scorelines and therefore contains all of the scorelines within any possible set. This includes every scoreline from the perfect set, where one player wins every point, to all of the possible scorelines leading to a tie break where every point has a Point Value of 50%.

The leverage landscape for each of these possible sets is significantly different. Some sets are full of break points and tight games, while others are filled with routine holds. Looking at the PV graph, some games can be as flat as Salt Lake from beginning to end. Others can be studded with multiple peaks resembling the Rocky Mountains.

To compare performances across sets or matches, we normalize the accumulated PV, add it to the median and express it as a percentage, to give us the Probability Optimization (POp) metric.

The POp metric is a clean measure of how well a player converted the leverage they faced, and therefore its value ranges from 0 to 100 percent.

0%: the opponent converted or won all available Point Value
50%: the player broke even and converted or won 50% of the Point Value, the same as the opponent
100%: the player converted or won all available Point Value

POp is not a psychological metric. It’s not a guess about tactics. It’s not a story. It’s a value‑realization metric - a measure of how close the player came to GTO Tennis Play given the points they actually faced.

The grey set of points that is graphed in the Tennis Point Value tool is the POp value for each point in the set.

For display purposes, these points are not centered over the 50% value on the Y axis. Instead, the points are shifted down and centered on the Y value of 10 (the dark grid line) to make it easier to see the relationship between the POp graph line and the changes in the Point Value graph line.

The higher above the dark grid line the POp values are, the better the player is at playing the Point Value probabilities. When the points fall below the dark grid line then the opponent is the one doing a better job of playing the Point Value probabilities.

Note that even though the graph is shifted, you can still hover over each POp point in the graph to see the actual POp value for that point. The overall POp value for the set is provided in the Win/Loss alert at the end of the set.

Although the tool plots POp for every point in the set, the most important value is the last value. This represents the players overall POp value and is a measure of how well the player traversed the leverage landscape of the entire set.

The Djokovic Effect: GTO Tennis in the Real World

The best way to take advantage of the leverage landscape is for players to Optimize their Tactical and Strategic decision-making in sync with the PV of each point.

To demonstrate this, in Playing Tennis Like "Deep Blue" Part 2 we created a Point Value model based on slightly different parameters.

In this second model, we maintained the same average winning percentage across all points but changed the winning percentage into a range or spectrum of 3% higher to 3% lower than the original value and then applied this range of winning percentage based on the PV of each point. You can think of the 3% lower value as a player's Good tactics, and the 3% higher value as a player's Great tactics, and so the spectrum of choices is from Good to Great.

We then ran this model to simulate millions of games with the Great, or highest winning percentages, applied to higher PV points, and the Good, or lower winning percentages, applied to lower PV points.

This slight tactical change, where a player uses their most effective or Great tactics and strategies on the higher PV points, and less effective but still Good tactics and strategies on lower PV points, increased their set winning percentage by 8% from 50% for the original model to 58% for the modified model.

If you want a real‑world example of a player that actively varies their tactics like this, have a look at Novak Djokovic.

He doesn’t magically “rise” on big points. He doesn’t summon clutch out of thin air. He clearly has learned, either consciously or intuitively, to play Great on "Big" points by:

utilizing his most decisive tactical and strategic patterns
pushing his mindset and execution quality to the highest level
and surprising his opponents with choices that force them to make low‑PV decisions

This has allowed Djokovic over the years to capture a disproportionate share of high Point Value points. When you do that, and do it well, you are going to win a lot of matches.

The purpose of the Probability Optimization (POp) metric is to quantify this behavior. It measures how well a player optimized their probability of winning the match.

We plan, as time allows, to run one or more of Djokovic's historical matches through the Tennis Scoreline Probabilities tool and document the analysis to show how he:

stretches his tactical and strategic repertoire into a spectrum going from Good to Great
and holds his greatest tactics and strategies close to his chest, only bringing them out at the highest PV moments

We like thinking of Probability Optimizing (POp) as the Djokovic Effect since he is extremely good at it, but it is not folklore and it is not just for pros. The Point Value model and the Tennis Scorelines Probability tool reveal it to be a value optimization strategy that any tennis player at any level can strive to maximize.

How Coaches and Players Can Use This Today

All of this will be more valuable once the 2026 BallBOPPer and BallBOPPer App are available, but right now with the Tennis Scoreline Probabilities tool, players and coaches can:

enter a real set point by point and see how the player performed on "Big Points"
or generate a random set and compare how Point Value plays out across a wide variety of scorelines
and then analyze the leverage landscape revealed by the Point Value graph for each set
and finally get the calculated player's POp metric for a particular set to see how well the player optimized their probability of winning that set

This gives you:

a map of where the match actually turned for or against the player
a measure of how well the player handled the big points
a way to evaluate tactical discipline under pressure
a clean, reproducible metric of performance quality

Traditional stats tell you what happened. The PV and the POp metric tells you how much it mattered.

Soon we plan to add the ability to run historical ATP/WTA Grand Slam matches through the Tennis Scoreline Probabilities tool, assuming we can find the time to implement this in between testing and demonstrating the 2026 BallBOPPer prototype.