SMA & EMA

Yes, and it comes from how the recursion is seeded rather than from the standard EMA formula. The loop sets `ema = p[0]` before any smoothing is applied (line 17), so Price t-4 enters at full weight; every later price, starting with Price t-3, gets discounted by (1 − alpha) for each step still ahead of it. At the default alpha of 0.4, that makes Price t-4's effective weight 0.1296 versus Price t-3's 0.0864 — the oldest price outweighs the second-oldest.

No — SMA(5) is a plain average of all five prices (line 15) that never references alpha anywhere in its formula. Alpha only shapes EMA(5), where a higher value shifts more weight onto Price t (the newest observation) and less onto the earlier prices in the recursive blend.

Price t, the newest observation, at the calculator's default alpha of 0.4 — it receives the recursion's last smoothing weight, alpha itself (0.4), directly and undiscounted, versus Price t-4's seeded weight of (1 − alpha)⁴ = 0.1296. That is not true for every alpha, though: Price t's weight is alpha while Price t-4's is (1 − alpha)⁴, and those two cross around alpha ≈ 0.2755. Below that (e.g. the smallest selectable alpha, 0.05), Price t-4 actually carries more weight than Price t — 0.05 versus 0.95⁴ = 0.8145.

SMA(5) treats all five prices identically regardless of order, so it simply reports the flat average; EMA(5) weights recent prices more heavily through the recursive blend (line 19), except for the oldest price's outsized seed weight noted above. Whenever the series is trending rather than flat, that different weighting makes EMA(5) diverge from SMA(5) — more so as alpha rises.