BoxingP4P

P4P Rankings, Statistics and Boxing news.

Methodology

How we compute the metrics used on this site and how they feed into the Pound-for-Pound (P4P) ranking. All formulas are applied consistently to every boxer in the database. For a short overview of how the P4P score is built from its components, see the rankings page (BoxingP4P Metric card).

Level of Opposition (LoO)

What it measures: How strong a boxer's opposition has been, using each opponent's record at the time of the bout (no look-ahead). We compute a LoO score per bout, then take the average over the chosen window (full career or last 10 fights).

Per-bout LoO. For each bout, we use the opponent's wins and losses before that bout (w_j and ℓ_j). If the opponent had no prior bouts, we skip that bout in the aggregate. Otherwise:

LoObout,j=100×wjwj+ℓj+K\text{LoO}_{\text{bout},j} = 100 \times \frac{w_j}{w_j + \ell_j + K}

Shrinkage constant K. The constant K in the denominator is a proprietary smoothing parameter that prevents extreme scores when the opponent has very few fights. As the opponent's record grows, the formula converges toward their true win rate.

Aggregate LoO. Average of per-bout LoO over all informative bouts in the window:

LoO=1∣I∣∑j∈ILoObout,j\text{LoO} = \frac{1}{|I|} \sum_{j \in I} \text{LoO}_{\text{bout},j}

Versus One

What it measures: All opponents are collapsed into one "super opponent" whose record is the sum of all opponents' wins and losses at the time of each bout. Result is a 0–100 score (win rate of that composite opponent).

Bouts in career order. Over informative bouts only (where w_j + ℓ_j > 0), after bout i:

VersusOnei=100×∑j=1iwj∑j=1iwj+∑j=1iℓj\text{VersusOne}_i = 100 \times \frac{\sum_{j=1}^{i} w_j}{\sum_{j=1}^{i} w_j + \sum_{j=1}^{i} \ell_j}

Career Versus One = value at the last bout.

Risk Factor

What it measures: How often a boxer takes fights significantly above or below their own usual level of opposition over time. Signal: per-bout LoO (v_1, …, v_n) in career order.

For each bout i ≥ 2, compute mean and standard deviation on the window v_1, …, v_i:

μi=1i∑k=1ivkσi=1i∑k=1i(vk−μi)2\mu_i = \frac{1}{i} \sum_{k=1}^{i} v_k \qquad\qquad \sigma_i = \sqrt{\frac{1}{i} \sum_{k=1}^{i} (v_k - \mu_i)^2}

With a proprietary threshold k, each bout i is classified:

  • High-risk bouts High-risk if v_i ≥ μ_i + k·σ_i
  • Low-risk bouts Low-risk if v_i ≤ μ_i − k·σ_i
  • Neutral otherwise

Raw score and weighted raw (μ_global = mean LoO over all boxers' bouts, used so that risk is scaled by typical opposition level):

raw=highCount−lowCountclassifiedBoutsweightedRaw=raw×μglobal100\text{raw} = \frac{\text{highCount} - \text{lowCount}}{\text{classifiedBouts}} \qquad \text{weightedRaw} = \text{raw} \times \frac{\mu_{\text{global}}}{100}

We clamp weightedRaw to [−1, 1], then map to 0–100 with 50 = neutral. The UI displays this score minus 50, so neutral appears as 0 (positive = more risk-taking, negative = more conservative).

RiskFactor=50+50×clamp(weightedRaw, −1, 1)\text{RiskFactor} = 50 + 50 \times \text{clamp}(\text{weightedRaw},\ -1,\ 1)

Risk Factor is a metric that is relative only to the boxer's own career path (their μ_i, σ_i over time). It therefore cannot be used in the P4P ranking calculation, which compares boxers to one another; it is shown for information only.

Weight Cutting

What it measures: Career weight cutting from height and bout weight classes. Natural weight (kg) is defined as follows (height in m; the reference BMI depends on the bout's weight class, see table below):

NaturalWeight=heightm2×BMIref\text{NaturalWeight} = \text{height}_m^2 \times \text{BMI}_{\text{ref}}

Per bout with valid weight-class limit:

cutj=100×NaturalWeight−LimitjNaturalWeight\text{cut}_j = 100 \times \frac{\text{NaturalWeight} - \text{Limit}_j}{\text{NaturalWeight}}

A proprietary reference BMI is assigned to each weight class, calibrated to the typical body composition for each division. These values are not disclosed.

Raw career score (mean over all bouts):

WeightCuttingraw=1N∑j=1Ncutj\text{WeightCutting}_{\text{raw}} = \frac{1}{N} \sum_{j=1}^{N} \text{cut}_j

This value can be negative (fighting at or above natural weight). For P4P, the raw percentage is transformed using a proprietary neutral-band mapping to a 0–100 scale (50 = neutral). The UI shows the raw percentage for transparency.

Weight Cutting is an experimental metric: height (BMI) and weight-class data per boxer are not yet very stable or precise in our sources, so results may still evolve as data quality improves.

Age + Activity

What it measures: A 0–100 score combining Recency (how long ago the last fight was) and Activity (average fights per year over the window).

Recency — based on years since last bout. The longer a boxer has been inactive, the lower the recency score. Scoring thresholds are proprietary.

Activity — based on fights per year over the window. More frequent fighters score higher. Scoring thresholds are proprietary.

The two components are combined using proprietary weights:

AgeActivity=clamp(α×recency+β×activity, 0, 100)\text{AgeActivity} = \text{clamp}(\alpha \times \text{recency} + \beta \times \text{activity},\ 0,\ 100)

For inactive boxers, Age/Activity is excluded from the P4P formula.

Win Rate

What it measures: Success rate against opposition (distinct from LoO, which measures strength of opposition).

WinRate=100×winswins+losses+draws\text{WinRate} = 100 \times \frac{\text{wins}}{\text{wins} + \text{losses} + \text{draws}}

If no recorded bouts, WinRate = 0.