How to read a Pronoisys table

CEP is a qualitative analysis of tennis matches, running since 2007. A 6–3 set tells you nothing about how the set was won — cruising, or saving ten break points and a set point. CEP evaluates the how, and it weighs every result by the quality of the opposition:

“The value of the defeated gives glory to the winner — and the value of the winner gives proportional value to the defeated.”

Every number below comes from that idea. The exact formulas are proprietary; what you need for reading is what each value measures and which direction is good.

Reading basics

• Mirror layout. Player A occupies the left half, Player B the right, labels sit in the middle. Every value on the left has its twin on the right — you always read side against side.
• Three main categories, and they are read together: Final Winner (the pre-match estimate), Per-set estimated scenario (how the match may unfold, set by set) and Tie-break dynamics. When they disagree, the risk of miscalculation rises — that warning has been part of CEP since the beginning.
• Colors do the first pass. Green favours that player, red counts against; deeper shade = stronger signal. Yellow boxes carry the headline values. Gold text is labels, never data.
• Grey numbers are sample sizes. A brilliant value from 2 matches is a rumour; the same value from 40 matches is a fact.
• Surface row above, all-courts row below. Nearly every block shows a surface-specific line over a general line. CEP keeps five separate evaluation histories per player — clay, hard, indoor, grass, and all courts together.

The Final Winner table base

hand

Each player's dominant hand — context for the serve/return picture and for the opponent's hand block further down.

core

The player's medium-term value: computed from the last 10 matches on this surface (surface row) and the last 20 on all courts (general row). There are two core computations per section — one median-based, one average-based — because the two can disagree in telling ways, which is why you'll also see two yellow value boxes.

running

The short-term value: built from the player's matches in the tournament currently in progress. This is the "how are they this week" number, against core's "who have they been lately".

RM — Reliability Meter

How consistent the player's evaluations are, on a 1–10 scale. It answers: can you trust the core value, or does this player swing wildly match to match? 7.0+ is strong, 5.5–7.0 moderate (caution), below 5.5 the player is unpredictable — treat their other values with suspicion.

timeous

A freshness grade for the data itself: how recent the player's last five evaluations are. A+ means all five within the last 15 days; the scale degrades through A, B, C to D (fewer than five evaluations in 4.5 months). A player returning from a long absence carries a poor timeous grade — their values exist, but they're stale.

final winner value (the yellow boxes)

The headline output: the model's separation between the two players, expressed as the minimum price at which that side would represent value. Two boxes because there are two core computations (median and average) — when both point the same way, the picture is firm; when they split, the median/average disagreement is itself a warning.

the percentages between the yellow boxes (EM)

The Effectiveness Meter: each player's actual recent success rate, as a reality check on the model values. It exists to catch a specific species — the player whose evaluations look good but who keeps failing to close matches out. Good values + weak EM = mentality question.

return / service form

Separate form ratings for the two halves of the game — surface line first (last 10 matches plus a time window), general line beneath (last 20 plus a time window). The window is about two months for the match and 1st set, and longer in the later sets, where players have fewer matches. Read the crosses: A's serve against B's return and vice versa. Two strong servers point to few breaks and long sets; strong returners against shaky serves point to a broken rhythm. A player better than their opponent on both serve and return is the strongest single configuration this block can show.

Baseline (the +/- number under each form value)

The player's hold rate (under service) or break rate (under return) against the average for this exact set situation, in points, over all point-by-point history with older matches fading: a match from 2 years ago counts half. Plus = above average, minus = below. The form value above it shows recent trend, Baseline shows what the player usually does. Read serve against the opponent's return: the higher one is green.

the blue number in each top corner

The estimated number of breaks made by the player on that side, in that set (in the whole match in the Final Winner block), from the form values and Baseline. An average, not a forecast of one exact score: a single set can easily differ by one break.

all courts / {surface} cep evaluations

The last three matches on any court and the last three on this surface — date, opponent, and the CEP value produced that day. This is the raw evidence behind core and running. Crucially, a win can carry a negative evaluation — an unconvincing victory over a weak opponent is recorded as exactly that, and it should make you cautious.

h2h by CEP

The head-to-head in values rather than raw wins: average CEP value produced against this exact opponent, general and surface, with counts. A player can lead a rivalry 3–1 yet consistently produce weaker values in it — this row catches that.

30d / 15d / Week Court Time (hrs)

Hours on court in the last month, fortnight and running week. Fatigue matters in a footwork sport: heavy recent load is a warning, while near-zero hours can mean rust. Always read it relative to the opponent's number.

opponent's hand

Historical CEP values split by opponent handedness, with counts. Some players genuinely underperform against lefties — here is where it shows, before you find out the hard way.

Running CEP chart base

Both players' running values plotted over recent months on a single axis. Crossovers, convergence and one-way trends are visible at a glance — the fastest way to tell whether a running number is a blip or a trend.

Criterias & historical CEP values add-on

How each player has historically performed under the specific conditions of this match. Each row is one criterion: left and right show each player's average CEP value under it, grey counts give the sample (general + surface).

city / tournament

Average value at this exact tournament. Some players own certain venues; some never fire at them.

country / continent

The same idea, widening the net — travel, conditions and comfort effects live here.

combined rows (level + draw + venue)

The strictest filters: performance in matches most similar to this one — same tour level, same draw size, same place.

avg cep value win / loss

What the player's values look like on winning days versus losing days. A narrow gap = their level barely drops in defeat (dangerous even when losing); a wide gap = when they're off, they're really off.

serve / return (tournament) and (h2h)

The serve/return form picture recomputed inside this tournament only, and inside this head-to-head only.

Raw stats — big points / serve / set-1 / decider add-on

A different species from CEP values: raw percentages from the last 12 months of point-by-point data, both players, general + surface splits. Direction noted per row:

BP saved % (serve)

Break points saved on own serve. Higher = hard to break when it matters.

BP converted % (return)

Break points converted on return. Higher = takes the chances they create.

2nd-serve pts won %

Points won behind the second serve — the serve's floor when the first serve misses. Higher is better.

double faults / serve game

DF frequency. Lower is better.

SET-1 serve pts % / SET-1 return pts %

Raw serve and return points won in first sets only — who actually starts matches well, in points rather than model values.

tie-break POINTS won %

Points won inside tie-breaks, all sets — a bigger, more stable sample than tie-break win/loss records.

clutch (big pts − avg, pp)

Big-point performance minus the player's own ordinary level, in percentage points. Positive = raises their game on the points that decide; negative = shrinks below their own standard.

nerve: decider DF − set1 DF

Does the serve tighten under maximum pressure? Double faults in deciding sets versus set one. Lower (or negative) is better.

slow-start (s1 pts − match, pp)

First-set level versus whole-match level. Negative = slow starter — they may still win matches, but set one is where they bleed.

DECIDER win %

The record when a match goes the distance, with the number of such matches.

OPP POOL — median rank (12-mo)

The median ranking of the opponents behind all of the above. Lower = tougher pool. This is the quality-control row for the entire table: numbers earned against strong opposition travel; numbers earned against a weak pool flatter the player. When one side's stats were built against far weaker opponents, discount them.

Per-set scenarios add-on

The section traders live in: instead of one match-level estimate, each scenario gets its own full evaluation block — form, Baseline, timeous, RM, core, running, values, last-3 evaluations, h2h — built only from matches in that situation.

1st set estimation

Both players evaluated on first-set performance history only. Read exactly like the Final Winner block; the question is who takes the opener.

2nd set estimations — W1 / L1

W1 = won the 1st set: the player's 2nd-set behaviour after taking the opener. L1 = lost the 1st set: their 2nd-set behaviour when trailing. The table cross-matches the branches — A's W1 values face B's L1 values, and vice versa — so both possible states of the match are pre-analysed before it starts. Front-runners and comeback players separate cleanly here.

3rd set estimations — W1L2 / L1W2

W1L2 = won the 1st, lost the 2nd; L1W2 = lost the 1st, won the 2nd. The two roads into a decider, each with its own history: who closes it out after letting a lead slip, and who finishes the comeback they started.

Tie-break dynamics add-on

tie break dynamics

The full evaluation block computed on tie-break performance overall — who each player has been in breakers, value-wise, not just win/loss.

1st set tie break

Restricted to first-set breakers — pairs naturally with the 1st-set estimation.

decisive set tie break

Restricted to final-set breakers — the single highest-pressure situation the sport offers.

Live scores base

A live board sits at the top of the dashboard sidebar, refreshing every few seconds with every in-play match that has a table. It's built for reading the match against its per-set scenarios in real time.

click a live match → land on the playing scenario

Opening a live match doesn't just load its table — it scrolls you straight to the section for the set being played right now, and draws a frame around the exact branch that is live: in a 2nd set, the W1/L1 row of whoever took the opener; in a 3rd set, the W1L2 / L1W2 path (teal frame for the 2nd set, purple for the 3rd). No hunting — the relevant scenario is in front of you.

ppH2H — point-by-point head-to-head

A small ppH2H tag sits in the bottom-right of every live card. One tap opens — right here in the panel — the break battles that actually decide sets: how often each player holds from 15-0, saves from 0-15 down, holds from deuce, serves out the set and comes through tie-breaks — one player's serve cross-matched against the other's return, side by side. Every number carries its sample size, with a pool-quality row so you know who earned it against tougher opponents. The scenario follows the live score (set 1, then the set-1 winner's angle, then the 1-1 decider), so it's always the situation being played right now.

colors

Dark red = live (men), a lighter red = women. Blue = just starting (fewer than two games played — score data too thin to act on). Grey = suspended: no points for 3+ minutes (hover to see for how long); it stays on the board for up to 4 hours and turns red again the instant play resumes. Green = just finished — the final score stays visible for 5 minutes, then the match leaves the board.

what rises to the top

The board self-sorts by urgency: just-finished results first (green, with a flash), then tie-breaks in progress, then matches whose set just ended (they pulse for a minute), then everything else. So the top of the list is always where something is being decided.

the * next to a games number

Marks who is serving: the asterisk sits on the server's games in the current set (e.g. 6-4, 3*-2 — the left player is serving at 3-2).

tie-break alerts

When any set reaches 6-6 the match jumps to the top wearing a gold TB badge, and a distinct urgent trill sounds — you will know a breaker started without looking.

sounds & the 🔊 toggle

A four-note bell rings when a set ends or a match finishes; the rising trill is tie-breaks only. Each event chimes exactly once. The 🔊 button in the top bar switches sounds on or off (remembered per browser). Note: browsers unlock audio only after your first click anywhere on the page — click once and you're set.

yellow "updated" markers

Separately from live scores, a sidebar entry turns yellow when its table was regenerated since you last opened it — fresh numbers worth a re-read.

How the room reads a table

• Read it whole. The strongest pictures are concordant: Final Winner, the per-set paths and the tie-break blocks leaning the same way. When the categories disagree, the risk of miscalculation rises — treat a loud value in a quiet table as a headline, not a case.
• Reliability before value. A big separation from a player with RM below 5.5 or a C/D timeous grade is built on sand. Check RM and timeous before believing anything else.
• Respect the samples. Grey counts first, values second. Thin data on one side usually means qualifiers, injury gaps or a level jump — which is itself information.
• Check who the values were earned against. The OPP POOL row and the named opponents in the evaluations tell you whether the numbers met real resistance.
• Surface first, general as the sanity check. When the two rows disagree, understand why before trusting either.
• Nothing here is a tip. Pronoisys shows the evidence; the conclusion — and the responsibility for it — is yours.