MIDI Drum Velocity vs Note Chance: Which Makes a Hit Softer—or Skips It?

Keep position, presence, velocity, receiver response, and timing separate

The MIDI Association describes Note On with a note number and velocity; velocity is normally used to control relative amplitude. That does not promise a fixed decibel change. The receiving patch decides what a velocity value does, so two drum sounds can react differently to the same input. Apple’s Step Sequencer documentation likewise notes that velocity usually controls attack loudness but may control another mapped parameter.

One useful inference follows from the sources, but it must remain an inference: standard Note On is documented with note number and velocity, while Note Chance appears as a feature of individual sequencers. Do not assume a probability setting is a portable field inside every MIDI handoff. Preserve the source project, and separately capture one playback result when an exact result matters.

Build A0 as a deterministic 16-step loop

Use one bar of straight sixteenth-note steps and four short, distinct sounds. Do not add swing, timing randomization, note repeat, round-robin changes, velocity processing, or bus effects during the test. The kick, snare, and closed-hat rows are anchors. The final percussion row is the only candidate for variation.

A0 contains 12 anchor events and four optional P events, or 16 voice-events per loop. Give every P event one documented, nonzero middle velocity that the receiver clearly registers. Record four loops into a playback/output capture downstream of the probability decision—not the source clip’s stored-note list. Each captured loop must show K on 1/9, S on 5/13, H on every odd step, and P on 4/8/12/16. Save or duplicate this control before changing anything.

Use the probability endpoints as wiring tests

Before listening to 50%, prove that you selected the right row and the right parameter. Set only the four P events to 0% for four loops. The downstream playback capture must retain exactly 12 anchor events per loop and contain no P Note On events. Then set those same P events to 100%. The four P events must return on every loop, restoring 16 total events. Their positions and velocities must match A0.

1. 0% gate: P is absent; K, S, and H remain complete. If anchors vanish, the selection or probability group is wider than intended.
2. 100% gate: P appears four times per loop. If it does not, solve mute, routing, mapping, or capture problems before discussing randomness.
3. Value gate: P’s captured velocity at 100% matches A0. A probability edit must not silently become a velocity edit.
4. Position gate: all played events remain on their original steps. Turn off any variation tool that also changes timing or note length.

These endpoints are deterministic controls. They test the setup; they do not evaluate the musical usefulness of probability. Only after both the 0% and 100% controls pass should you choose an intermediate value such as 50% for the observation run.

Isolate each dimension, then combine and return: A0 → V → C → B → A1

Create V, C, B, and A1 as independent copies of A0 rather than editing one version into the next. The arrow is only a presentation order. Compare A0↔V to isolate velocity, A0↔C to isolate chance, V↔B and C↔B to inspect the combined version from each one-variable side, and A0↔A1 to prove the return. This keeps A0 available and prevents V→C or B→A1 from masquerading as one-variable edits.

For V, choose two nonzero values that your current patch audibly distinguishes, then write the actual values in your notes; there is no universal “ghost” or “accent” number. Alternate them across P4/P8/P12/P16 while leaving every event present. For C, restore the single A0 velocity before lowering chance. This prevents a quiet surviving event from being mistaken for a skipped event.

B is optional and comes last. It demonstrates a strict order of operations in the evidence: first ask whether a P event exists on that pass; only then inspect the velocity of a P event that exists. Never enter a zero or invented velocity for an absent event. Finally, A1 must reproduce A0’s event positions, counts, and values—not merely sound “close.”

Observe 16 loops, but do not demand 32 hits from 64 opportunities at 50%

Ableton’s official FAQ explicitly warns that 50% Note Chance does not mean every second hit, and states that ordinary same-chance notes are calculated individually. Native Instruments describes intermediate per-note probability as being re-rolled on each pass in Maschine 3.7. Those statements support repeated observation, not an exact short-run quota. Other tools may document grouping or conditions differently.

Duplicate the loop rows through 16 or record an output event list downstream of the probability decision. The hard requirements are 192 anchor events across 16 loops, unchanged anchor positions and values, no P events in the 0% control, and all 64 P opportunities played in the 100% control. The 50% run has no “must equal 32” pass condition: log the played outcomes while every other dimension remains controlled. A finite run need not contain both present and absent outcomes. If all 64 are present or all are absent, recheck the endpoint controls and extend observation; do not call the engine broken or fair from that sample alone.

If you need “P16 on exactly every fourth loop,” probability is the wrong contract. Use a documented deterministic condition such as recurrence, first-loop/last-loop logic, or write the event into an explicit four-bar pattern. Apple documents Chance separately from Recurrence, First Loop, and Last Loop. Call the result conditional or arranged—not random.

Name the first invariant that fails

Listen only after the event audit names the dimension. A very soft sample can be inaudible while its MIDI event still exists; a missing event can leave reverb or another layer audible after it. Raw MIDI and dry audio answer different questions. Keep both when the boundary between sequencer and instrument is the suspected layer.

Map the concepts to documented tools, not shared labels

• MIDI Association: Note On is described with note number and velocity, and the receiver normally uses velocity for relative amplitude. It does not prescribe this article’s probability layer or a universal drum response.
• Ableton Live: the Live 12 manual exposes separate Velocity and Chance editors and describes a velocity deviation range. Ableton’s Live 11 FAQ calls its randomized range “Velocity Chance,” defines Note Chance as whether a note plays, and rejects the “every second hit” interpretation of 50%.
• Logic Pro: Apple documents Chance and Velocity as separate Step Sequencer edit modes. Its step inspector distinguishes stochastic Chance from deterministic Recurrence, First Loop, and Last Loop conditions.
• Maschine 3.7: Native Instruments distinguishes probability applied to existing notes from probability used during random note generation. Its per-note probability is re-evaluated on passes, while the documented Lock function can freeze a chosen result as fixed, editable MIDI.

Do not translate a name without translating its behavior. “Probability” might refer to writing notes, deciding playback, choosing one member of a group, or changing a parameter. Run the 0%/100% endpoint test, inspect actual events, and write down whether the operation is live, grouped, captured downstream, or frozen by a documented lock operation.

Preserve both the rule and one captured result

A probability-enabled source project and a captured playback result are not the same artifact. The project preserves the rule that can produce new passes. Only a MIDI/audio capture recorded downstream of the probability decision—or a product-specific lock/freeze operation explicitly documented to resolve it—can be treated as one fixed outcome. Do not assume an ordinary MIDI export or project reopen resolves or preserves Chance. If another collaborator must hear the exact same fills, hand off the captured MIDI or audio as well as the editable source, then verify event count, positions, velocities, sounds, and loop boundaries.

1. Name the source: include the sequencer and version, and state whether P uses independent, grouped, or conditional playback.
2. Capture enough context: begin one bar before the varying loop and end after the following downbeat so the boundary can be checked.
3. Keep A0: retain the deterministic reference beside the probability version; do not overwrite the only control.
4. Reopen: reload the source project and the separately captured result, then compare the first and last bars with the ledger.
5. Label the result: say “downstream MIDI capture,” “tool-documented locked MIDI,” or “rendered audio” instead of implying the probability rule itself became portable.

For pad practice, V can become a deliberate dynamic exercise because every event stays present. C is an arranging or playback variable, not proof of finger control. If the goal is to perform an optional fill reliably, write the chosen realization into a deterministic practice loop before judging timing or coordination.

Velocity and Note Chance FAQ

Does lowering Note Chance make the hits that remain quieter?

No such result follows from chance alone. On a pass where the note plays, its velocity and the receiver’s mapping determine the response. Test presence in MIDI before judging loudness in audio.

Why did 50% produce three hits in one section and none in another?

Because 50% is a probability, not an alternating schedule or a short-run quota. First confirm exact behavior at 0% and 100%, then observe multiple loops without demanding an exact half.

Why did changing velocity alter timbre more than level?

The receiving patch can map velocity to a documented parameter other than level, or ignore it. Record the incoming value and describe the observed receiver response instead of treating velocity as a decibel setting.

Should kick and snare always stay at 100%?

That is the control policy for this test, not a universal arranging rule. Protecting anchors makes the difference between presence and value observable. A later arrangement may vary structural hits intentionally in a separately named version.

Is Note Chance part of standard MIDI Note On?

The cited MIDI overview documents Note On with key number and velocity, while the chance behavior cited here belongs to specific sequencers. It is therefore safest to treat portability as unproven until the destination reproduces a controlled test; this is an inference from the sources, not a new MIDI specification claim.