PennAhead

How accurate the estimates are

Every estimate PennAhead shows is checked against the track NJ TRANSIT announces. These numbers come straight from that log, from September 30, 2026 to November 17, 2026: 5338 departures with an announced track over 49 days. Nothing here is hand-typed.

Headline
90%+ estimates

When the board showed a track at 90%+, it was right 3775 of 3790 times (99%). 3790 of 5338 departures (71%) had a 90%+ estimate before the track was announced.

Over time
Right track, by day

0%50%100%9/3011/17

All estimates on the board, by service day.

Over time
Right platform, by day

0%50%100%9/3011/17

Estimates whose track or its island partner (Tracks 1–2, 3–4, …) was the one announced, by service day.

Coverage
How often there is an estimate

5337 of 5338 departures (99%) had an estimate on the board before NJ TRANSIT announced the track. The rest were announced before PennAhead showed an estimate, usually because estimates from past track patterns appear only within 20 minutes of departure.

By confidence
Right track, by what the board showed

0%50%100%3775/379090%+16/16HIGH13/21MED375/1510LOW
  • 90%+right 3775 of 3790 (99%)
  • HIGH (75–89%)right 16 of 16 (100%) · small sample
  • MED (60–74%)right 13 of 21 (62%) · small sample
  • LOW (under 60%)right 375 of 1510 (25%)
  • All estimatesright 4179 of 5337 (78%)

Right platform: 4367 of 5337 (82%). Tracks that share an island platform count as one, so an estimate of Track 7 when the train left from Track 8 still sent you to the right stairs.

Lead time
How early a 90%+ estimate appeared

For correct 90%+ estimates, the board first showed that track a median of 20 minutes before NJ TRANSIT announced it, and 30 minutes before the scheduled departure (3775 departures).

By method
Where the estimate came from

  • Train already on a platformright 3185 of 3199 (99%)
  • Past track patternsright 390 of 1533 (25%)
  • Train on the approach to a platformright 583 of 584 (99%)
  • Train just moved off its platformright 15 of 15 (100%) · small sample
  • Train at the far end of a platformright 6 of 6 (100%) · small sample

By line
Right track, by line

0%50%100%1867/2467NEC1053/1372NJCL1032/1225M&E166/197MOBO61/76RARV
  • NECright 1867 of 2467 (76%)
  • NJCLright 1053 of 1372 (77%)
  • M&Eright 1032 of 1225 (84%)
  • MOBOright 166 of 197 (84%)
  • RARVright 61 of 76 (80%)

Misses
Recent wrong estimates

Of 1158 wrong estimates, 188 still pointed to the right platform (16%): the train left from the other track on the same island, so the same stairs.

  • #3247 NJCL, Nov 17 1:43 PM
    Estimated Track 3 (44%), left from Track 2
    Wrong platform
  • #3239 NJCL, Nov 17 11:23 AM
    Estimated Track 1 (21%), left from Track 7
    Wrong platform
  • #6317 M&E, Nov 17 9:22 AM
    Estimated Track 3 (36%), left from Track 4
    Right platform (Tracks 3–4)
  • #3223 NJCL, Nov 17 7:44 AM
    Estimated Track 1 (27%), left from Track 13
    Wrong platform
  • #3705 NEC, Nov 17 6:26 AM
    Estimated Track 8 (28%), left from Track 13
    Wrong platform
  • #6209 MOBO, Nov 17 5:53 AM
    Estimated Track 1 (26%), left from Track 15
    Wrong platform
  • #3701 NEC, Nov 17 5:32 AM
    Estimated Track 13 (23%), left from Track 2
    Wrong platform
  • #3805 NEC, Nov 17 1:22 AM
    Estimated Track 4 (28%), left from Track 6
    Wrong platform
  • #3803 NEC, Nov 17 12:43 AM
    Estimated Track 4 (16%), left from Track 2
    Wrong platform
  • #3201 NJCL, Nov 17 12:05 AM
    Estimated Track 4 (24%), left from Track 10
    Wrong platform
  • #3293 NJCL, Nov 16 10:23 PM
    Estimated Track 12 (22%), left from Track 11
    Right platform (Tracks 11–12)
  • #3887 NEC, Nov 16 9:06 PM
    Estimated Track 3 (23%), left from Track 8
    Wrong platform
  • #3875 NEC, Nov 16 6:47 PM
    Estimated Track 10 (19%), left from Track 4
    Wrong platform
  • #3373 NJCL, Nov 16 5:56 PM
    Estimated Track 12 (20%), left from Track 10
    Wrong platform
  • #3871 NEC, Nov 16 5:50 PM
    Estimated Track 12 (27%), left from Track 13
    Wrong platform

Improving
Where the estimates need to get better

The weak spot is plain to see in these numbers. When the train is already on or near its platform, the estimate is almost always right. When PennAhead has only past track patterns to go on, it's right far less often, and those are the estimates that need the most work.

PennAhead collects this data every day: every estimate it shows, where trains were before their tracks were posted, and the track NJ TRANSIT finally announced. The results are checked daily, and every change to how estimates are made is measured against this record and kept only if it helps. This page updates as that work lands.

How it's measured
What counts

For each NJ TRANSIT departure, the estimate counted is the last one on the board before the track was announced, and it's right if it matches the announced track. Canceled trains are left out. The board shows estimates of 90% or more as 90%+; the results on this page are exact. NJ TRANSIT can change a track at any time, so always check the station displays.

Last updated November 17, 2026