Deadlock Hero Balance Predictions
Rem buff/nerf predictor prototype
Generated from the Rem predictor DuckDB, recent patch history, and a three-stage prediction model.
Prediction Model
Rem Buff/Nerf Watch
This page is the hero-side version of the item prediction screen. The target is whether a hero looks more likely to be buffed, nerfed, or left alone in a future patch. This first page is Rem-only so the predictors can be reviewed before expanding to every hero.
Best read: this is now a three-stage experiment. It first asks whether Rem is likely to change, then which broad mechanic would be selected, then the direction of that selected mechanic. Later stages are conditional and must not be read as standalone certainty.
- Stage 1
- Will Rem receive any balance change in the next patch?
- Stage 2
- If Rem changes, which mechanic family is most likely to be selected?
- Stage 3
- If that family is selected, does its patch direction look like buff, nerf, or mixed/rework?
- Future learning
- Weekly snapshots preserve hero stats and ability values over time. They are enabled as model features only after changed and unchanged patches both have adequate coverage.
Three-Stage Next-Patch Forecast
1
Will Rem change?
11.9%
Estimated chance of any Rem balance change. No-change estimate: 88.1%.
Disclaimer: Experimental. This is based on one hero's patch history and has not been proven accurate — Stage 1 and Stage 2 don't yet beat a simple "no change" guess on holdout data. Treat every number on this page as a rough signal, not a confirmed forecast.
Stages 2 and 3: target and direction if Rem changes
| Mechanic | Selected if changed | Direction | Buff if selected | Nerf if selected | Mixed/rework if selected | Combined priority |
|---|---|---|---|---|---|---|
| Naptime | 3.0% | Mixed Or Rework | 0.0% | 13.4% | 86.6% | 0.4% |
| Vitality | 2.9% | Mixed Or Rework | 0.0% | 32.9% | 67.1% | 0.3% |
| Spirit | 2.6% | Mixed Or Rework | 0.0% | 32.9% | 67.1% | 0.3% |
| Weapon | 2.5% | Mixed Or Rework | 0.0% | 32.9% | 67.1% | 0.3% |
| Tag Along | 1.4% | Mixed Or Rework | 0.0% | 16.4% | 83.6% | 0.2% |
| Lil Helpers | 1.1% | Mixed Or Rework | 0.0% | 17.8% | 82.2% | 0.1% |
| Pillow Toss | 0.8% | Mixed Or Rework | 0.0% | 23.9% | 76.1% | 0.1% |
Combined priority multiplies Stage 1 by Stage 2 to help sort the watchlist. It is not a calibrated probability that the named mechanic will change.
Validation and Data Readiness
Every stage uses older patches for training and newer patches for testing. The current patch calendar contains 138 update dates through 2026-06-12.
| Stage | Test rows | Balanced accuracy | Macro F1 |
|---|---|---|---|
| Stage 1 - hero changedWhether Rem receives any balance change. | 28 | 50.0% | 0.472 |
| Stage 2 - mechanic selectedWhich mechanic family is selected if Rem changes. | 196 | 50.0% | 0.491 |
| Stage 3 - conditional directionBuff, nerf, or mixed/rework after a mechanic is selected. | 3 | 100.0% | 1.000 |
Performance feature gate: withheld. Historical pre-patch win-rate coverage is 0.0%; the required coverage is 80.0%. Performance features are withheld because missingness is too sparse or too strongly associated with whether Rem changed.
The weekly collector now stores all hero performance, base stats, and ability properties without overwriting prior snapshots. Once enough unchanged and changed patches have comparable coverage, performance and ability-strength features can enter training without leaking the answer.
Rem Patch Changes
24 Rem patch-change rows are included from the recent patch window.
| Patch | Predictor | Direction | Before | After | Notes |
|---|---|---|---|---|---|
| 2026-05-22 | naptimeability:naptime | Rework | Naptime now provides +30% damage reduction in the base ability (T3 still adds +50%) | ||
| 2026-05-22 | pillow_toss_charge_timeability:pillow_toss | Nerf | 9s | 8s | Pillow Toss charge time reduced from 9s to 8s |
| 2026-05-22 | pillow_toss_t3ability:pillow_toss | Buff | +90 | +100 Damage | Pillow Toss T3 increased from +90 to +100 Damage |
| 2026-05-22 | reduced_lockout_period_after_ejecting_from_tag_along_from_1s_to_0_3sability:tag_along | Nerf | Tag Along from 1s | 0 | Reduced lockout period after ejecting from Tag Along from 1s to 0.3s |
| 2026-05-22 | tag_alongability:tag_along | Rework | Tag Along fixed specific abilities/items that could kill Rem while he's napping | ||
| 2026-05-22 | tag_alongability:tag_along | Rework | Tag Along now allows friendly auras to affect Rem | ||
| 2026-05-22 | tag_along_burst_healability:tag_along | Buff | Tag Along burst heal now also applies to secondary allies that you jump to | ||
| 2026-05-22 | tag_along_canability:tag_along | Rework | Tag Along can now be cast through veils | ||
| 2026-05-22 | tag_along_ejectability:tag_along | Nerf | Tag Along eject no longer briefly restricts movement abilities and input | ||
| 2026-05-22 | tag_along_eject_isability:tag_along | Nerf | Tag Along eject is no longer affected by debuff resistance and now launches upwards slightly more to prevent hitting the ground | ||
| 2026-05-22 | tag_along_heal_duration_isability:tag_along | Rework | Tag Along heal duration is now displayed on the hud as well as the healthbar icon | ||
| 2026-05-22 | tag_along_lingering_heal_durationability:tag_along | Nerf | 3s | 2s (values rebalanced to be the same overall heal just faster) | Tag Along lingering heal duration reduced from 3s to 2s (values rebalanced to be the same overall heal just faster) |
| 2026-05-22 | tag_along_t2_barrier_and_item_duration_range_effectivenessability:tag_along | Buff | 25% | 35% | Tag Along T2 Barrier and Item Duration/Range effectiveness increased from 25% to 35% |
| 2026-05-22 | tag_along_t3ability:tag_along | Buff | 0.03 | 0 | Tag Along T3 now also increases Missing Health Heal from 0.03 to 0.05 and Regen Per Second from 0.66 to 1.0 |
| 2026-05-22 | tag_along_time_to_fly_to_your_ally_isability:tag_along | Rework | Tag Along time to fly to your ally is now 50% faster | ||
| 2026-04-30 | fixed_a_bug_with_naptime_that_allowed_you_to_attack_while_casting_itability:naptime | Rework | Fixed a bug with Naptime that allowed you to attack while casting it | ||
| 2026-04-30 | pillow_toss_recharge_timerability:pillow_toss | Nerf | 10s | 9s | Pillow Toss recharge timer reduced from 10s to 9s |
| 2026-03-06 | lil_helpers_trooper_base_healingability:lil_helpers | Nerf | 10/s | 8/s | Lil Helpers trooper base healing changed from 10/s to 8/s. The delay between heals has been reduced so the frequency of healing is more constant. (overall reduction) |
| 2026-03-06 | lil_helpers_trooper_healing_spirit_scalingability:lil_helpers | Buff | 0.125 | 0 | Lil Helpers trooper healing spirit scaling increased from 0.125 to 0.14 |
| 2026-03-06 | naptime_radiusability:naptime | Nerf | 20m | 17 | Naptime radius decreased from 20m to 17.0m |
| 2026-03-06 | naptime_sleep_durationability:naptime | Nerf | 4.25s | 4s | Naptime sleep duration decreased from 4.25s to 4s |
| 2026-03-06 | naptime_sleep_move_speedability:naptime | Nerf | 2m/s | 1 | Naptime sleep move speed reduced from 2m/s to 1.5m/s |
| 2026-03-06 | naptime_t2ability:naptime | Rework | "unstoppable and -35% Damage resist while channeling" | "+0 | Naptime T2 changed from "unstoppable and -35% Damage resist while channeling" to "+0.5s Sleep Duration and +3m Radius" |
| 2026-03-06 | naptime_t3ability:naptime | Buff | Naptime T3 now also grants +50% damage resist while channeling |
Hero Balance Predictions
Rem buff/nerf prototype
Forecasts whether Rem is likely to change next patch, using recent patch history and the Rem predictor DuckDB.