How Our Competitions Work — and How You Can Verify Them

Randomness in computing is surprisingly hard to do well. Software algorithms can appear random but are ultimately deterministic —given the same starting state, they produce the same sequence. That's not good enough for a fair draw.

Our platform uses a dedicated ID Quantique Quantis PCIehardware quantum random number generator. This device produces genuine randomness from quantum optical processes: it measures the behaviour of individual photons, which is fundamentally unpredictable by the laws of physics — not just "hard to predict", but provably without a hidden pattern. No algorithm, no seed, no back-door.

This hash is published as part of the result. Because SHA-256 is a one-way function, the only way to produce the same hash is to have the exact same inputs. The commitment proves the seed was fixed before selection began — it is mathematically impossible to swap the seed for a different one without the hash changing.

The PDF certificate issued after each draw contains the following fields:

• Commitment hash — proves the seed was locked in before selection. If you hold this hash from before the draw, you can verify the seed was not chosen after the outcome was known.
• Reveal seed — the raw quantum entropy. Feeding this into the verifier replays the entire draw from scratch and produces the same winners.
• Transcript hash — proves every operation happened in the recorded order. Changing any winner, or changing their sequence, breaks this hash.

Go

Python — no installation required:

JavaScript — run with Node.js:

If you omit --commitment-hash , --transcript-hash , or --winners , those checks are skipped and the tool simply prints the computed values.

Expected output:

=== Draw Verification ===
Draw ID:           summer-prize-draw-2026
Total Entries:     8500
Number of Winners: 5
Reveal Seed:       9f2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a0b1c2d3e4f5a6b7c8d9e0f1a2
Commitment Hash:   c8d4f3e9a2b1c0d9e8f7a6b5c4d3e2f1a0b9c8d7e6f5a4b3c2d1e0f9a8b7c6
Transcript Hash:   d7c6b5a49382f1e0c9d8b7a6f5e4d3c2b1a0f9e8d7c6b5a4938271f0e9d8c7b6
Discarded Samples: 0
Winners (0-idx):   7823,412,5901,3367,88

✓ commitment_hash  MATCH
✓ transcript_hash  MATCH
✓ winners          MATCH

Verification PASSED.

Python:

JavaScript:

Go:

Expected output:

=== Draw Verification ===
Draw ID:              summer-prize-draw-2026
Total Entries:        8500
Number of Winners:    5
Number of Alternates: 3
Reveal Seed:          9f2b3c4d5e6f7a8b9c0d1e2f3a4b5c6d7e8f9a0b1c2d3e4f5a6b7c8d9e0f1a2
Commitment Hash:      a1e4f7b2c5d8e3f096a2b1c4d7e0f3a6b9c2d5e8f1a4b7c0d3e6f9a2b5c8d1e4
Transcript Hash:      f3a2b1c0d9e8f7a6b5c4d3e2f1a0b9c8d7e6f5a4b3c2d1e0f9a8b7c6d5e4f3a2
Discarded Samples:    0
Winners (0-idx):      7823,412,5901,3367,88
Alternates (0-idx):   2204,6711,1055

✓ commitment_hash  MATCH
✓ transcript_hash  MATCH
✓ winners          MATCH
✓ alternates       MATCH

Verification PASSED.

Some draws display entries numbered from 1 (where 1 = the first entry in the pool). If the certificate says "1-indexed", add --one-indexed and pass both winners and alternates exactly as printed on the certificate — the flag applies to both lists:

Running the verifier and seeing Verification PASSED confirms three things:

1. The seed was fixed before the draw. The commitment hash matches, which means the 32-byte quantum seed was locked in — along with the draw parameters — before any winner was selected. No one could have seen the outcome and then chosen a different seed.
2. The exact draw algorithm was followed. The transcript hash matches, which means the recorded sequence of operations produced these exact winners and alternates. Swapping any winner or alternate, or reordering them, produces a different transcript hash.
3. The winners and alternates came from genuine randomness. The seed is 32 bytes of certified hardware quantum entropy. The verifier replays the selection from that seed and arrives at the same winners and alternates — meaning no alternative list is consistent with this seed.

Selecting a random integer in [0, max) by computing r % max on a uniformly distributed 32-bit value introduces a small bias unless max divides 2^32 exactly. Our implementation removes this bias with rejection sampling:

Only values below limit are accepted, so every outcome in [0, max) has exactly the same probability. The number of discarded values is reported on the certificate as Discarded Samples and is usually zero. For draws with up to 100 million entries, the expected discard rate is under 0.01%.