Algorithm Validation Report
NavCommand & IncCommand checked cell for cell against the U.S. Navy Diving Manual, Rev 7 Change A✓ Every value both apps use from Tables 9-4 through 9-9 matches the published manual. 6,384,403 automated checks ran through the real app code and screens, including an exhaustive grid of 4.56 million dives and a battery of what-if scenarios. None found an app error. The October process found 13 defects; all are fixed, and every check was re-run after the fixes.
Checked Against the Manual Itself
We extracted the air tables from the U.S. Navy Diving Manual, Revision 7 Change A, and compared them with both apps cell by cell. Values were compared directly, with no sampling.
| Table | What was compared | Cells | Result |
|---|---|---|---|
| 9-7 No-decompression limits | Every limit and group time, 10–190 fsw | 240 | ✓ Exact |
| 9-8 Residual nitrogen time | Every RNT cell, including the “**” and “†” cells | 384 | ✓ Exact |
| 9-8 Surface-interval credit | Every interval boundary for all 16 groups | 136 | ✓ Exact |
| 9-9 Air decompression | Every schedule 30–190 fsw: stops, ascent time, group, exceptional-exposure and mode lines | 338 | ✓ Exact |
| 9-4 Sea level equivalent depth | Every cell, including the cells the manual leaves blank | 318 | ✓ Exact |
| 9-4 Equivalent stop depths | Every stop at every altitude | 60 | ✓ Exact |
| 9-5 Group on arrival at altitude | All 10 altitudes | 10 | ✓ Exact |
| 9-6 Wait before ascent to altitude | Every group × every altitude increase, plus Notes 1–4 | 160 | ✓ Exact |
The Core Suites: 100,279 Checks, Zero Failures
Each suite loads the real app file in a headless browser and calls the same code a supervisor uses, or reads the answer off the screen. The results are compared with an independent reference model written from the manual’s rules.
Random and boundary dives in both apps: table row, residual nitrogen, equivalent single dive time, adjusted limit, deco decision, schedule row, total ascent time, actual stop depths at altitude, surfacing group and exceptional exposure. Covers sea level to 10,000 ft, warm and cold, clean and repetitive divers.
✓ 41,872 / 41,872Each cell of Tables 9-4, 9-6, 9-7, 9-8 and 9-9 is read back through each app’s own lookup code. Ascent time is checked against the stops on every schedule. Also covers surface-interval credit, altitude arrival groups, altitude rounding, actual stop depths, and oxygen limits for every nitrox mix.
✓ 35,453 / 35,453The full workflow (Dive 1, surface interval, Dive 2) is driven through the real planner screens. Group, stops, ascent time and warnings are then read off the display, the way a diver reads them.
✓ 7,289 / 7,2895,000 identical dives are planned in both NavCommand and IncCommand, and the two answers must match. Another 10,000 results, weighted toward no-stop dives, are checked against the reference.
✓ 15,000 / 15,000The 18-hour rule after a schedule with no repetitive group (9-9.1 item 9). No repetitive dive on an exceptional-exposure schedule (item 10). Correct exceptional-exposure flags. The decompression-mode note on every Table 9-9 row. The 300–1,000 ft altitude rule. The 18-hour window in each app’s dive log.
✓ 665 / 665To prove the suites would notice a real mistake, we planted 22 realistic bugs, one at a time. Examples: one mistyped stop time, one wrong Table 9-4 cell, a missing schedule row, the 18-hour rule removed, rock bottom without the minute at depth, 30 extra minutes of surface-interval credit. All 22 were caught.
22 caught · 0 missedWhat If the Day Doesn’t Go to Plan?
Real dives rarely run exactly as planned. These suites push both apps through overstays, short surface intervals, long multi-dive days, altitude, cold water, bad input and lost gas, and check every answer against the manual’s rules.
Every whole-foot depth from 1 to 190 ft, not just the table rows, so odd depths like 47 ft or 133 ft are covered. Each one ran at 100 bottom times, every altitude band from sea level to 10,000 ft, warm and cold, and clean or carrying Group C, H, M or Z. Each answer was compared field by field: table row, residual time, decompression, stops at their actual depth, ascent time and group.
✓ 4,560,000 / 4,560,000For a clean diver, one more foot or one more minute must never mean less decompression. 1,506,130 neighbouring pairs were checked. 16 pairs were flagged, and all come from the manual’s own rows. Example: cold at 50 ft / 74 min, the next longer schedule is the 95-minute row (4-minute ascent), while at 55 ft it is the 75-minute row (3 minutes).
✓ 0 app errors · 16 as publishedDays of 2 to 5 dives were written into NavCommand’s real dive log, at sea level and altitude. They include surface intervals of 2 minutes to 18 hours, decompression and exceptional-exposure dives. The diver’s group, residual time and go / no-go were then read at 17 moments after the last dive and 7 planned depths, and compared with an independent replay of the Rev 7 rules.
✓ 30,000 / 30,000IncCommand’s live log: a planned dive is started, and the diver leaves the bottom early, on time or up to 40 minutes late. In some runs no “left bottom” tap is made. The log must re-plan from the actual time, never come out better than the plan, and flag omitted decompression when the overstay needed stops that weren’t planned. 1,484 of the runs did.
✓ 6,000 / 6,000Rock bottom in both apps against an independent gas model. Covers 10–130 ft, no-stop and decompression dives, five cylinders, breathing rates from 10 to 40 psi/min (and none entered), sea level to 10,000 ft, salt and fresh water.
✓ 23,863 / 23,863The planned, +5 min, +10 fsw and both rows that each app shows under a plan, recomputed independently. Covers residual nitrogen, altitude and cold, including exceptional-exposure and no-group schedules.
✓ 80,000 / 80,000Team members with up to five dives in the last 60 hours, some exceptional exposure or no-group. Checks that the highest group of the last 24 hours (Table 9-6 Note 1) and the 48-hour exceptional-exposure wait set the time to clear, for mountain passes, commercial flights and sites above 8,000 ft.
✓ 10,000 / 10,000Every mix from 21% to 40% oxygen, at every depth the app will offer, at six altitudes, in salt and fresh water. ppO₂ never passes 1.4, and the limit is never set needlessly shallow.
✓ 61,626 / 61,6266,400 altitude-arrival cases: a team that drove up 0–15 hours ago, combined with a logged or manual starting group (Table 9-5). 98 odd entries, such as blanks, negatives, “1e9”, “45ft” and non-Latin digits, with no crashes and no wrong plans. 4 reload tests of altitude, cold and fresh-water settings. 3 checks for page errors across the runs.
✓ 6,505 / 6,50513 Defects Found, All Fixed
The point of testing is to find problems before a diver does. These are the defects this process found. Each is fixed and covered by a check that would fail if it came back.
Verified, and Where the Edges Are
- ✓ Single and repetitive air dives on every Table 9-7 / 9-9 row (IncCommand plans to 130 fsw, the OSHA SCUBA limit; NavCommand 30–190 fsw)
- ✓ Surface intervals under 10 minutes treated as one dive
- ✓ Altitude from sea level to 10,000 ft: Tables 9-4, 9-5 and 9-6
- ✓ Cold / strenuous: next longer schedule
- ✓ Exceptional exposure, no-group and decompression-mode warnings
- ✓ Nitrox oxygen limits at sea level and altitude, in salt and fresh water
- ● Air dives deeper than 190 fsw. All are exceptional exposure, and the apps do not plan them
- ● Surface decompression and in-water oxygen schedules. The apps plan in-water air and show the manual’s mode note
- ● Helium (mixed-gas) diving
- ● DiveCommand (Bühlmann ZHL-16C). Its validation is being re-run under the same method
How the Tests Were Run
The manual is the reference, not the app. The air tables were extracted from the Rev 7 Change A PDF. The reference model loads Tables 9-4 and 9-9 straight from that extraction, so app data is never compared with itself. Tables 9-5 through 9-8 were diffed cell for cell against the PDF. Where the text extraction was ambiguous, the PDF page was checked by eye.
An independent model of the rules. The reference is a separate Python implementation of the Rev 7 procedure, written from the manual’s rules, not from the app code: table selection, residual nitrogen, the 10-minute rule, altitude correction, cold / strenuous and exceptional exposure.
The real apps, in a real browser. Both production app files run in headless Chromium through Playwright. Tests call the same functions the planner uses, and the end-to-end suite reads the answers off the screen.
Testing the tests. A suite that never fails proves nothing on its own. So 22 realistic bugs were planted one at a time: 14 in the table and rule logic, and 8 more in the what-if areas (rock bottom, fresh water, overstay, the dive log, flying, altitude arrival and nitrox). The suites had to catch every one. They caught all 22.
What-if scenarios are generated, not hand-picked. Dive days, overstays and gas scenarios are drawn at random from realistic ranges with fixed seeds, so every run can be repeated exactly. The exhaustive grid covers every whole-foot depth, so nothing between the published rows is skipped.
Where the manual is silent. In a few places Rev 7 doesn’t say exactly what to do. In those cases the apps take the conservative reading and say so in their “how we calculate” pages. One example: a schedule with no repetitive group gets the 48-hour exceptional-exposure wait before altitude, because Table 9-6 has no group for it.
See exactly how each number is worked out: NavCommand: How We Calculate · IncCommand: How We Calculate. The test suite is available on request for agency safety officers and reviewers: contact us.