When a split moves your level

The short answer

A split changes the unit a price is quoted in, not the stock, so a level built before it can sit at a multiple of the new price. In GEXfreak’s rebuilt daily rows, the wall in force on split day was still in the old unit: one session is built in (walls are built after the close), and a second came where the rebuilt data had no split-day row.

AAPL’s 4-for-1 split on 31 August 2020 is the example. Its close as traded went from 498.90 to 128.85, while the stock itself rose 3.3%. A wall is a strike, so it stays in the old unit until a new one is built.

We did not check what the live Prism by GEXfreak chart draws on split day; the first bullet below is how you check.

What to do with this

• On split day, compare each wall with the price on your screen before you use it. The split date and ratio are in the company’s notice. A wall several times the price, like AAPL’s 500.00 call wall over a 128.85 close, is in the old unit. Divide it by the split ratio (500.00 ÷ 4 = 125.00) and use the result only to know roughly where the wall was: no stop, target or alert on it until a wall built after the split appears (step 4). For the share splits here that took one to two sessions.
• Check your own lines, alerts and open orders. A price you set before a split is in the old unit unless your broker converted it, and this page did not measure what brokers do.
• If you back-test a level or count how often price reaches it, leave split day and the next session out, and look back by date, not by rows. A row with no wall makes “five rows back” reach further than five sessions (in our rows, five rows before 1 September 2020 is 24 August 2020, six sessions earlier), and an old-unit wall puts any distance to it off by the ratio.
• Do not read a converted wall as a prediction. The converted call wall matched the first fresh one five times out of five and the put wall two times, which is too few to rely on. It would take about 100 logged splits to see a gap smaller than 10 percentage points, so wait for the fresh row.

How we tested

GEXfreak read daily closes for 12 stocks from 26 September 2016 to 31 December 2024, raw and adjusted for splits, with each stock’s recorded split list. For the four stocks that had an event we took the last level row before it and the first row after it.

To judge whether a wall carrying over is unusual, we counted how often those four stocks’ 6,671 level rows (from 1 May 2018) kept the same wall a set number of sessions apart, with no split between them. That happened for 71.1% of call walls and 69.7% of put walls one session apart, and 60.4% and 59.0% two sessions apart. Four splits are compared with the two-session rates and NVDA in 2024 with the one-session rates, and the events are treated as independent. On that basis, keeping all five call walls has a 9.5% chance and keeping two or fewer put walls has a 29.8% chance. A difference was not detected, and five events could only have shown a very large one.

IBM on 4 November 2021 is in the figure but is not a split. Holders received Kyndryl Holdings shares, so earlier prices are multiplied by 1000 ÷ 1046 (0.956) and nobody’s IBM share count changes. Its raw move was −4.9% and its adjusted move −0.6%. It has 28 sessions with no level row, so its walls say nothing here; for a spin-off, read the company’s notice.

Check it yourself with Prism and a spreadsheet

Splits are rare, so rehearse first: in the figure choose AAPL, open “Table, method and data source”, and use its rows for the worked lines. For a real split, keep one sheet per wall, call and put, headers in row 1: row 2 is the session before the split, row 3 is split day, then one row per session until column G reads “new unit”.

The steps, and what a small log can say

• 1. Check whether your broker chart adjusts. Open AAPL on a daily chart and read the 28 August 2020 close: 498.90 means the chart shows prices as traded, 124.73 means it is adjusted for the split, so turn the split adjustment off (if it cannot be turned off, multiply an adjusted close by the ratio). Closes can differ slightly from the minute-bar closes used here (more if adjusted for dividends). Write the ratio in column B on every row. Row 2: B2 = 4, C2 = 28 August 2020, D2 = 498.90. Row 3: B3 = 4, C3 = 31 August 2020, D3 = 128.85. Rows 4 and 5 (Sep 1 and Sep 2) also take B = 4, D4 = 134.20 and D5 = 131.22.
• 2. Work out the moves. H3 =D3/D2-1 gives −74.2% and I3 =D3/(D2/B3)-1 gives +3.3%. A big raw fall with a small adjusted move is consistent with a split; confirm it on the company’s notice.
• 3. Read the wall off Prism at each open, starting on split day, and write it in column E; if what Prism shows differs from the figure’s table, note it. On the figure’s table the wall you see at an open is the newest row built before it: the Aug 31 and Sep 1 opens both show Aug 28’s row, so E3 = E4 = 500.00 for the call wall (450.00 for the put wall). The Sep 2 open shows Sep 1’s row, so E5 = 125.00. Stop when column G reads “new unit”.
• 4. Run the unit check. F3 =E3/D3 = 500.00 ÷ 128.85 = 3.88, and G3 =IF(ABS(F3-B3)<ABS(F3-1),"old unit","new unit") reads “old unit” because 3.88 is closer to the ratio 4 than to 1. Fill F3:G3 down to row 5. Row 4 is 500.00 ÷ 134.20 = 3.73, also old unit; row 5 is 125.00 ÷ 131.22 = 0.95, “new unit”. Rows reading “old unit” are stale sessions (2 here). At a live open the close is not there yet, so divide by the price on your screen. In the recorded rows it flagged all 18 call and put walls on the nine stale split sessions as old unit, at 3.0 to 9.9 times the close.
• 5. Convert and compare. J3 =E3/B3 = 500.00 ÷ 4 = 125.00, and on the first “new unit” row K5 =IF(ABS(E5-J3)<0.005,"same","differs") reads “same” because E5 is 125.00. The put block reads “differs”: 450.00 ÷ 4 = 112.50, but the first fresh row had 125.00. This page holds no option-adjustment notices, so it cannot say why.
• 6. Read your log at its size. Divide the count of “same” in K (=COUNTIF(K:K,"same")) by the count of rows that read “same” or “differs”, and compare it with the ordinary rate for your wall and gap from How we tested (59.0% for a put wall over two sessions). The 95% range is, for n = 20 splits and p = 0.59, =1.96*SQRT(0.59*(1-0.59)/20) = 0.22.

Limits

• Six events (five splits and one spin-off adjustment) in four of twelve stocks that GEXfreak chose and that were all still listed at the end of 2024. The count is for these stocks, not the market. There was no reverse split; the unit check is untested on one, so check a reverse split by eye.
• The walls are read back from GEXfreak’s end-of-day store of saved daily levels, rebuilt after the close from stored option data, not a live feed or our research rebuild at five-minute marks, which has no row on any of these six days, so nothing here says what an intraday map drew then; this page does not say which expiries each wall includes. Up to 27 March 2023 the rebuilt data has no wall for split day: all five events in that period (four splits and the IBM spin-off) have no split-day row. Where a reason was recorded (IBM’s 28 missing rows) it is that no option-chain rows were found; the recorded rows do not say why. The one later event (NVDA, 10 June 2024) had a row and one stale session, so it is the only clean case. What a live Prism by GEXfreak chart draws on split day was not measured; the log above is how you check it.
• Only call and put walls were measured. The daily, weekly and monthly flips, the expected move and the prior-day high and low were not, so check each against the new price the same way. IBM’s adjustment alone accounts for 4.4% of its 4.9% raw fall. Adjusted prices are adjusted for splits only, in the share count of 31 December 2024.
• The closer-to-the-ratio line in the unit check is ours and was not tested against other choices. The log sizes treat the ordinary rate as exact. Nothing here is a trade or a forecast.