Market Internals for Index Futures: Reading TICK, ADD and VOLD Without Fooling Yourself

TICK, ADD and VOLD are not three views of the same thing. They use three different reference prices, carry three different amounts of memory, and answer three different questions. Almost every dollar lost on them comes from one of two errors: assuming the three should agree, or treating a reading as a trade signal when it is only context for a setup you generated some other way. This piece defines each one precisely, shows why two of them routinely contradict each other, and works the arithmetic of the only use that holds up.

What do TICK, ADD and VOLD actually measure?

Each one is defined by its reference price. TICK compares each issue's last trade to the immediately preceding trade in that same issue. ADD and VOLD compare each issue's current price to the previous trading day's close. That single difference explains nearly all of their behaviour.

The NYSE TICK is, in dxFeed's plain wording, "the difference between the number of stocks which have an uptick and stocks with downtick." If 1,842 issues in the counted universe last traded on an uptick and 604 last traded on a downtick at the same instant, TICK is 1,842 minus 604, or +1,238. Those two counts do not sum to the full listed universe, because trades printed at the same price as the previous trade (zero-plus and zero-minus ticks) are generally excluded. TICK is not a trade counter and not a volume measure. It is a headcount of directional participation at one instant.

That plus and minus tick vocabulary predates charting software. It comes from US short sale regulation: under the tick test the SEC applied to short sales before repealing it, a short sale was permitted on a plus tick, meaning a price above the immediately preceding sale, or on a zero-plus tick, meaning a sale at the last price when that price is higher than the last different price, and was prohibited on minus and zero-minus ticks. TICK counts that same taxonomy.

ADD is the net issue count against yesterday's close. dxFeed defines advanced issues as how many assets in the counted universe currently trade higher than their previous trading day close price, with declined issues as the mirror. One correction to a common description: the intraday ADD most futures traders watch is not cumulative. The classic Advance-Decline Line sums net advances across days. Intraday ADD is a point-in-time net count that rises and falls all session as individual issues cross back and forth over their own flat line.

VOLD is that same bucket split, weighted by shares. dxFeed's advanced volume is "the today's trading volume of all assets in the user defined universe which currently trade higher than their previous trading day close price", and VOLD is that figure minus its declined mirror. Read the construction carefully: the bucket assignment is per issue, not per trade. An issue's entire session volume sits in whichever bucket its current price puts it in. Some vendors instead compute volume traded on upticks versus downticks, which behaves differently, so confirm which construction your feed uses before you interpret it.

Why does cash equity breadth matter to a futures trader?

Because the index future and the cash constituents are wired together by arbitrage, and breadth describes the constituents on the other end of that wire. ES and NQ are futures on cash indices. The index level is a function of its constituent stock prices. Any persistent gap between the future and the basket is closed by desks that trade both sides.

Regulators have described that plumbing directly. The joint CFTC and SEC staff report on the market events of May 6, 2010 attributed part of that day's cascade to cross market arbitrageurs who absorbed E-mini sell pressure and passed it into the equity market, buying E-mini contracts while simultaneously selling products like SPY or selling individual equities in the S&P 500 Index. The wire runs both ways, and it runs fast.

So the path is a loop with named parts. Constituent stocks print. Those prints set the index level and, at the same instant, produce every breadth count: uptick versus downtick feeds TICK, above versus below yesterday's close feeds ADD and VOLD. Arbitrage desks watch the gap between ES and the basket, then close it by sending orders into the ES book, which moves the future, which changes the gap again. Internals let you watch the constituent end of that loop directly instead of only the single price that falls out of it.

The practical consequence: index price tells you the net result, internals tell you how it was produced. A 10 point ES move built on 2,000 issues rising together is a structurally different event from the same 10 points produced by four mega-caps while most issues fall. The chart is identical.

What is your feed actually counting?

Almost certainly NYSE-listed issues, and that matters more than most traders realise. The default $TICK, $ADD and $VOLD symbols on most platforms cover the NYSE universe. Listing venue, unlike index weight, is a stable fact about a company: Apple, Microsoft, Nvidia, Amazon, Alphabet, Meta, Broadcom and Tesla all list on Nasdaq, while Berkshire Hathaway, JPMorgan, Exxon Mobil and Walmart list on NYSE. Several of the heaviest S&P 500 constituents therefore sit outside NYSE breadth entirely. Check current weights with S&P Dow Jones Indices before assuming how much of the index your breadth screen covers.

That produces a repeatable trap. NYSE breadth can be strongly positive while the S&P 500 falls, because a few Nasdaq-listed giants can outweigh hundreds of NYSE advancers in a float-adjusted market cap index. The trader sees ADD at +900 against a red ES chart and concludes the data is broken. It is not. It is measuring a universe that excludes the names doing the damage. For NQ the problem is definitional rather than statistical: Nasdaq-100 membership requires a Nasdaq listing, so NYSE breadth covers none of that index. If you trade both contracts, this compounds the divergences covered in ES, NQ, YM and RTY correlation risk, where index composition is the reason the four contracts stop moving together.

Three feed-level details to confirm rather than assume. Vendors including dxFeed compute these counts over a user defined universe, so two traders can watch "the same" indicator and see different numbers. Symbol naming is not standard either: most retail platforms use $ADD for the issue count difference and $VOLD for the volume difference, while dxFeed's own indicator glossary applies the ADD label to the difference between advance and decline volume. And update cadence belongs to your feed, not to the indicator; the widely repeated claim that TICK is computed on a fixed six second cycle is not something exchange or vendor documentation confirms, so ask your data provider what yours does.

Why do ADD and VOLD disagree?

Because one counts companies and the other counts shares, and those two things concentrate differently. When they disagree, that is the information, not a fault.

Work a concrete case. Advancing issues: 1,700. Declining issues: 1,100. ADD is 1,700 minus 1,100, or +600, comfortably positive. Up volume: 1,020,000,000 shares. Down volume: 1,320,000,000 shares. VOLD is 1,020,000,000 minus 1,320,000,000, or negative 300,000,000 shares. Divide it out: volume per advancing issue is 1,020,000,000 / 1,700 = 600,000 shares, and volume per declining issue is 1,320,000,000 / 1,100 = 1,200,000 shares. The average decliner traded exactly twice the shares of the average advancer. Decliners are 1,100 of 2,800 directional issues, or 39.29% by count, but 1,320,000,000 of 2,340,000,000 shares, or 56.41% by volume. More things went up. The money went the other way.

TRIN, the Arms Index, compresses that into one number: (advancing issues / declining issues) divided by (up volume / down volume). On these figures that is (1,700 / 1,100) / (1,020,000,000 / 1,320,000,000), which is 1.5454... divided by 0.7727..., exactly 2.00. As fractions, (17/11) divided by (17/22) equals 22/11 equals 2. A TRIN above 1 means volume is concentrated in decliners relative to what the issue counts alone imply. Same disagreement, expressed as a ratio.

VOLD also moves in a way ADD structurally cannot, and this is where new users decide their data glitched. Under the per-issue construction, an issue's whole session volume lives in one bucket. Suppose one name has traded 40,000,000 shares while sitting below yesterday's close, so all 40,000,000 shares sit in the down bucket. It then prints one tick above yesterday's close. The entire 40,000,000 changes buckets: down volume falls by 40,000,000, up volume rises by 40,000,000, and VOLD moves 80,000,000 shares on a single print. ADD moves from (A - D) to (A + 1) - (D - 1), which is exactly +2. Applied to the numbers above, VOLD steps from negative 300,000,000 to negative 220,000,000 while ADD ticks from +600 to +602.

One issue crossing one price produces a 40 million to one difference in optical magnitude between the two series. That is why VOLD looks jumpy and ADD looks smooth, and why a VOLD "breakout" can be one large-cap name crossing its flat line rather than any change in market character. Feeds computing uptick and downtick volume do not produce this flip, which is another reason to confirm your vendor's definition.

The three misreadings that cost money

All three come from asking an indicator to do a job its construction does not support.

One: an extreme TICK print is an automatic reversal. Extreme prints come from basket and program activity hitting hundreds of issues in the same instant, which is exactly what index replication and cross-market arbitrage produce. On a genuine trend day that flow repeats, so extremes cluster in the direction of the trend all session instead of marking one exhaustion point. Fading the first +1000 on a strong up day is how traders learn this personally. It is a consequence of the mechanism rather than a published exchange statistic, so measure it in your own data before relying on it. The useful version is the inverse: on a trending day, note where TICK stops going, not where it reaches. A market that refuses to print below negative 400 on pullbacks is telling you who is absent from the selling.

Two: reading internals on a timeframe they do not support. TICK carries no state, so a value expires the moment enough constituent issues print again. Charting it as 1 minute or 5 minute OHLC bars shows the extreme reached inside that interval, and that extreme may have lasted a fraction of a second. A 5 minute bar high of +1,200 describes a condition that no longer exists and possibly never persisted beyond one refresh of the feed. ADD and VOLD are anchored to a fixed reference and do survive being charted. TICK largely does not. Cumulative TICK, summing values through the session, gives it artificial memory, but it is a platform or script construction rather than exchange-published data, so two platforms will not agree on its value.

Three: using internals as a standalone entry trigger. No reading of ADD, VOLD or TICK specifies an entry price, a stop or a target. They contain no level structure. The setup has to come from something that produces all three: a value area edge, a session high retest, a VWAP reclaim, an order flow and DOM read at a specific price. Internals then raise or lower your conviction in that setup. Reversing the order, hunting for a price setup because ADD looks strong, converts a context tool into a rationalisation engine for trades your plan never authorised.

When are internals actually live?

Only while the US cash equity market is open, which is a minority of the futures day. The NYSE core session runs 9:30 a.m. to 4:00 p.m. ET per NYSE's published hours, so 6.5 hours. The CME Globex trading day for equity index products runs from a 5:00 p.m. CT open to a 4:00 p.m. CT close the next day, a 23 hour window that also contains a short daily halt; verify current hours and contract specs with CME Group before relying on them. 6.5 divided by 23 is 0.2826, so internals are live for about 28% of that window. The other 16.5 hours, 71.7% of it, has no breadth data at all. Anything you traded overnight, you traded without them.

Two boundary cases catch experienced traders. NYSE publishes 1:00 p.m. ET early closes on a handful of days a year, and its current calendar lists July 3, 2026 for the observed Independence Day, November 27, 2026 for the day after Thanksgiving, and December 24, 2026. On those days internals go dark at 1:00 p.m. while equity index futures keep trading, and a trader running on an internalised 4:00 p.m. boundary spends the afternoon reading a frozen series. A 3.5 hour cash session is 3.5 / 23, or 15.2% of the futures day. Re-check the calendar each quarter, because the dates move with the weekday.

The opening window is the other one. NYSE's auction documentation states that orders "can be entered and cancelled until the security is opened by the DMM, which may occur after 9:30 a.m.", so the constituent universe does not arrive all at once. For the first minutes of the session, breadth counts are computed over a partial, still-filling set of issues. Early ADD and VOLD readings are structurally incomplete rather than merely noisy, and they drift as the rest of the universe opens.

A worked reading: internals as a sizing dial

Here is the only defensible use, with the arithmetic done. Price generates the trade. Internals set the size.

The setup, produced without internals: long ES on a pullback that holds above the prior session's value area high, stop 6.00 index points below entry, first target 12.00 points for 2R. ES is $50 per index point, so risk is 6.00 x $50 = $300 per contract and the target pays 12.00 x $50 = $600. Use MES at $5 per point for granularity, where ten MES equal one ES, so risk per MES is 6.00 x $5 = $30.

Internals aligned: ADD positive and rising, VOLD positive, TICK printing higher lows and refusing to go below negative 400 on pullbacks. Take 10 MES. Risk is 10 x $30 = $300, and the 12 point target pays 10 x 12.00 x $5 = $600.

Internals conflicting: identical price setup, but ADD is +600 while VOLD is negative 300,000,000, the exact divergence worked through above, with TRIN at 2.00. Broad participation, but the shares are trading in the decliners. Take 5 MES. Risk is 5 x $30 = $150 and the target pays 5 x 12.00 x $5 = $300. Same entry, same stop, same target. The only variable internals touched was exposure.

Internals do not create edge. They tell you which slice of an existing edge you are standing in.

That is an expectancy claim, so here is the arithmetic, using illustrative numbers you must replace with your own trade log. A base setup over 100 trades wins 45 at 2R and loses 55 at 1R: expectancy is (0.45 x 2) - (0.55 x 1) = 0.90 - 0.55 = +0.35R per trade, or +35R total. Now split by alignment. Suppose 60 of those trades had internals aligned and won 55%: (0.55 x 2) - (0.45 x 1) = 1.10 - 0.45 = +0.65R per trade. The remaining 40 must then have won 30%, since 0.60 x 0.55 + 0.40 x 0.30 = 0.33 + 0.12 = 0.45, which reconstructs the base 45% win rate. Those 40 trades have expectancy (0.30 x 2) - (0.70 x 1) = 0.60 - 0.70 = negative 0.10R. Check the parts against the whole: (60 x 0.65R) + (40 x negative 0.10R) = 39R - 4R = 35R, matching 100 x 0.35R exactly. Internals added no edge. They located a negative-expectancy subset inside a strategy that was already positive. That is worth a great deal, and it is a different claim from "internals make you money."

When adding internals is the wrong move

If your base strategy is not yet executed consistently, adding three data series will make things worse. This is the honest position and it disqualifies a lot of readers.

The costs are real and mostly unbudgeted. Breadth and index data carry separate market data entitlements at most vendors and the fees change, so verify current costs with your platform or data provider. Screen space is finite. Attention is finite. Every added series adds decision latency at the moment a scalper can least afford it. Put a number on it: if three extra series add 1.5 seconds of hesitation to an entry, and average adverse excursion on ES over 1.5 seconds is 1.00 index point, that is 1.00 x $50 = $50 per contract per trade. Six trades a day is 6 x $50 = $300 per contract per day in pure decision latency. The filter has to be worth more than that before it pays. Substitute your own measured slippage, because those inputs vary by product, size and time of day.

The realistic failure mode is not that internals are wrong. It is that they become a supply of reasons for trades the plan did not authorise. Add them only after the base strategy is already consistent, add them as a sizing modifier first, and never as an entry trigger.

One more limitation, specific to copied trading. Internals live on the lead trader's screen, not in the copier. A copier replicates fills, not context. If a lead uses internals to cut size by half, follower accounts see a smaller order and never see why, and any follower running different risk parameters translates that intent differently. Internals are also irrelevant to strategies copied on non-index products such as crude, gold or FX, and to anything running in the overnight session. A desk operating in those conditions gains nothing from adding these screens and should spend the attention elsewhere.

Frequently asked questions