McTrainer.Club PvP Benchmark Report August

Personal bests versus repeatable performance

Across sufficiently played player-mode combinations, the median gap between a personal best and the stable performance level was 38.4%.

The typical variation within the stable 10-run window was 36%.

A personal best answers: “What was my strongest individual run?”

A stable performance level answers: “What result can I reproduce consistently?”

This distinction matters because two players with the same personal best may have very different levels of consistency.

One player may reach that result regularly, while another may have achieved it only once.

How many runs does improvement take?

For this report, measurable improvement means a change of at least 5% between a player's first stable 10-run window and a later 10-run window.

The median player-mode history reached that threshold after 20 runs.

Overall, 50.6% of sufficiently long player-mode histories reached the threshold during the stored observation period.

This does not mean that every player will improve after exactly 20 runs.

Training mode, starting level, practice frequency, individual skill, and the number of available recorded runs all influence the result.

Example learning curve: Speed Anchor

Speed Anchor provides one of the clearest learning curves in the report.

The median time decreased from 573 ms during runs 1–5 to 350 ms among runs numbered 81 and above.

That represents an observed improvement of approximately 38.9%.

This curve should not be interpreted as a controlled experiment.

Later run groups naturally contain players who continued training for longer. This may introduce selection and survivor bias.

Even with that limitation, the result provides strong descriptive evidence that experienced Speed Anchor players produce substantially faster times than players in their earliest runs.

Other notable learning curves

Speed Anchor was not the only mode with a visible difference between early and later runs.

The analyzed data also showed:

• Totem Reaction: median time decreased from 2,165 ms to 1,028 ms, an observed improvement of 52.5%;
• Totem Speed: median time decreased from 1,700 ms to 858 ms, an observed improvement of 49.5%;
• Anchor Destruction: median time decreased from 9,652 ms to 6,598 ms, an observed improvement of 31.6%;
• First Hit: median score increased from 53.76 to 62.40, an observed improvement of 16.1%;
• Shield Break: median score increased from 68.13 to 81.55, an observed improvement of 19.7%;
• Aim Flicking: median KPS increased from 3.10 to 3.87, an observed improvement of 24.7%.

These comparisons describe differences between run-number groups. They do not prove that every individual player will follow the same progression.

Consistency differs significantly by mode

Some modes produce much more stable results than others.

Aim Flicking had a median stable variation of approximately 5.4%, while Small Flicking had a similarly low value of 5.9%.

By comparison, Speed Anchor had a median stable variation of approximately 74%.

This indicates that a player's recent Aim Flicking results are generally more repeatable, while Speed Anchor results can vary substantially between runs.

That difference may be caused by the structure of the modes themselves. A short timing mistake can have a large relative impact on a time-based result, while a longer aim test can average performance across more individual actions.

Ping analysis

A reliable ping comparison could not be produced for this edition because ping was not consistently present in the stored training-run metrics.

No conclusions about performance differences between latency groups are therefore included.

A future report may add this section once latency data is recorded consistently across training modes.

Exploratory comparison with Ranked PvP

Only 9 players met both requirements for this section:

• enough qualified training data to calculate an average training percentile;
• and at least five recorded Ranked matches.

Within this small sample, the Pearson correlation between average training percentile and Ranked skill rating was 0.576.

The correlation between average training percentile and Ranked win rate was 0.678.

Both values indicate a positive relationship inside this specific sample. However, nine players are not enough to support a reliable conclusion about the overall McTrainer.Club player base.

The Ranked comparison should therefore be understood as an early signal rather than a confirmed finding.

Future reports will be able to revisit this question with a larger sample.

What does your result mean?

A raw score becomes more useful when it is placed inside the wider performance distribution.

For example, a Sword Combo score of 32 is approximately at the 75th performance percentile of the qualified Sword Combo runs in this data set.

This means the result was equal to or better than approximately 75% of the analyzed values.

Why percentiles are useful

Every training mode uses a different measurement system.

Some modes measure:

• completion time;
• score;
• successful actions;
• attacks per second;
• or weighted accuracy.

A score of 50 in one mode cannot be meaningfully compared with a score of 50 in another mode.

Percentiles solve this problem by showing where a result sits within the distribution for its own mode.

They can answer questions such as:

• Is this result below or above the median?
• Is the player inside the top 25%?
• How far is the player's stable level from their personal best?
• Is the player's performance becoming more consistent?
• Has the player's percentile improved over time?

Limitations

This report has several important limitations.

• It is based on observational data and cannot establish causation.
• Players choose which modes to train and how often to train, creating self-selection bias.
• Later learning-curve groups contain only players who continued training long enough to reach them.
• The Ranked comparison contains only nine matched players and should be treated as exploratory.
• Current Ranked ratings were compared with training performance from the report period. This is not a historical before-and-after analysis.
• Ping data was not consistently available in training-run metrics.
• Different modes use different units and should not be compared directly using raw values.
• Some modes use a fallback score when their configured primary metric is missing from stored runs.
• Stored run histories may be limited by mode, meaning the oldest runs may not be available for every player.
• The report analyzes stored runs, not every attempt a player may have made outside the recorded system.
• Players with longer histories contribute more runs to the overall run-level distributions.

Conclusion

McTrainer can describe PvP training performance in a way that a traditional leaderboard cannot.

Instead of showing only the highest score, mctrainer.club can show:

• the full performance distribution;
• a player's percentile;
• the difference between peak and repeatable performance;
• consistency across recent runs;
• and how results differ as players complete more training.

The first McTrainer.Club PvP Benchmark Report analyzes 804,689 qualified runs from 5,514 players across 18 training modes.

The results show that personal bests are only one part of performance. Consistency, percentile position, and long-term progression provide important additional context.

Future editions can become even stronger as metric coverage improves, more Ranked matches are recorded, and additional data becomes available for latency and long-term player progression.