Vendor feeds as an input variable in a MACD evaluation

The MACD rule was held fixed

A MACD oscillator is used here as a fixed, unoptimized reversal rule that turns a repeatable chart condition into a testable long-or-short hypothesis. The historical exercise held one unoptimized MACD specification on one Russell 2000 sample to expose data-source effects. It was not designed to prove that a MACD reversal rule works.

The evaluation used a market-order fill, an at-market entry, so the comparison isolates data differences rather than order-type effects.

OHLC disagreements that look like hundredths

A five-year Russell 2000 sample from mid-2005 to mid-2010 can show vendor-to-vendor OHLC disagreements that look like hundredths on a single bar but do not cancel when summed. That pattern is accumulated tick error: tiny per-bar differences that look negligible on a single day but add into a material point total when summed across a multi-year sample.

In one two-vendor summation, closes were 52 points lower, opens 40 points higher, highs 65 points lower, and lows 48 points higher, with the signed gaps alternating rather than lining up in one direction. Range invariance, the idea that a day's high minus low might stay stable even when open and close prints disagree, is a check, not an assumption.

A 0.01 per-bar discrepancy accumulated over 1,257 trading days equals 12.57 index points. At 100 currency units per Russell 2000 point that is 1,257 units on a single contract path.

Two vendors can show long stretches of identical zeros in a difference sheet and still leave a non-zero close total, with most of one close gap concentrated on a single session.

Opens, rounding, and the exchange print

The originating index series supplied only closes for part of the window, to six decimal places, while vendor files typically carried two decimal places. That exchange print is a reference close, available at finer decimal precision than typical vendor files.

Vendors may not share the same definition of the open. Some record the first traded print, so equal-looking charts can still rest on different OHLC construction.

Signed outcomes under the same market-order rule

Under that fixed MACD market-order rule, commercial vendor series produced similarly negative outcomes, while the exchange series was the outlier and the only one near a non-negative result.

Across roughly 100 hypothetical trades the vendor-to-vendor P&L gap was only a few currency units, which is smaller than typical live bid-ask and fill variation.

Editorial: the small vendor-to-vendor P&L gap is why a standard oscillator rule can look consistent across commercial charts, even though the exchange series still shows that changing the feed can change the signed outcome.