1  What each column actually measures

2  The σ trap — why TIER lies

σ compares a candle to the 5 candles before it. When the tape goes quiet the standard deviation collapses, so an ordinary candle scores enormous σ. This is observed in your own data, not theory: a 1,885 lot print scored 68σ off a quiet [65,130,130,130,130] window — while the session's genuinely largest print scored only 3.1σ because the minutes around it were busy.

So a T3 badge in a dead stretch is worth less than a T1 badge in a busy one. Rank by the amber rows, not by TIER. Amber = biggest on the tape by actual size; dim amber = within half of it. Those two colours are the only size-honest ranking here.

3  Does a 70–80% share mean it's about to shoot up?

No. This was tested directly — every detected print split by share, then measured against VWAP 3 candles later, signed so "continued" means price moved the same way as the print's side. A real edge would sit well above 50%.

Not one symbol rewards following a high-share print. Concentration tells you a single participant was working one price. It says nothing about which way they intend to go next — a buyer filling 5,000 lots at one level is equally consistent with a breakout about to fire and with a level about to hold.

4  The one repeatable pattern: index absorption

The strongest consistent result in the whole dataset is a fade, not a follow. When an index print is big + one-sided (⋮≥75%) + concentrated (share ≥70%), price tends to go the other way:

Both indices, both horizons, all below 50% continuation — meaning roughly 60–64% of the time price reverses. That is the classic absorption footprint: someone is soaking up aggressive flow at a level rather than initiating through it. The aggressor is the one getting filled; the passive side is the one in control.

Absorption vs initiation — how to tell them apart live

• Absorption — big print, high share, price does not leave the bucket price over the next 2–3 candles. The level is being defended. Trade the level holding, not the print's direction.
• Initiation — big print, high imbalance, price closes away from the bucket price and keeps going. Only now is the print's side meaningful. Enter on the first pullback toward the bucket price, with that level as your invalidation.

The print alone never tells you which one it is. The next 2–3 candles do. This is exactly what CHART view is for — markers on the VWAP line so you can see whether size clustered at a level or drove through it.

5  Calibrate per symbol — "70%" is not one threshold

Share of candle depends on how wide each symbol's price bucket is (NIFTY 2.5 · BANKNIFTY 5 · CRUDE 5 · GOLD 10), so the same number means different things. On BankNifty 70% is below median. On Gold it is roughly the 73rd percentile. Use each symbol's own distribution:

A 70% share on Gold is genuinely unusual. A 70% share on BankNifty is a Tuesday. Never carry a threshold from one symbol to another.

6  Crude is the exception — it does trend

Crude is the only symbol where following size worked, and only on a 5-candle horizon — it decays to nothing by 10:

The strongest cell (70%, +13.4bps ≈ ₹12 on Crude) rests on only 10 samples — treat it as a hint to watch, not a system. The n=133 row at 59% is the one with enough data behind it to plan around.

7  How to actually use it

8  A workable routine

• Set MIN σ to 3 and leave it. Higher thresholds mostly filter out busy-period prints — which are the size-honest ones — while keeping quiet-stretch artifacts.
• Scan for amber rows first. Note the PRICE column on each; those buckets are your levels for the session.
• Switch to CHART. Check whether markers cluster at a level (absorption) or step price along (initiation). For Crude and Gold, turn GEX on — a big print landing on Call Resistance or Put Support is far more meaningful than one in open air.
• Check cum Δ in the summary line. If it disagrees with the print you like, take the smaller size or skip it. Watch especially for divergence — price flat while cum Δ climbs is a better tell than either alone.
• Wait 2–3 candles. You are not trading the print. You are trading what the print reveals once price responds to it.