After upgrading a regular gift, you get three attributes: backdrop, symbol, model. Each has its own rarity. Their combination determines the market price of the instance, and that dependency is non-linear. This article is a technical guide to the rarity system: how the tiers work, what probabilities sit behind them, how the market values them, and how to avoid mistakes when reading a lot card.
TL;DR
- Each of the three attributes (backdrop, symbol, model) has a rarity scale: usually common → uncommon → rare → epic → legendary.
- Drop probabilities for a specific attribute are visible on marketplaces in the collection card: from ~5-10% (common) down to 0.1-0.5% (legendary).
- Full lot rarity = product of three attribute probabilities. A “common × common × common” lot is typical; “epic × epic × epic” is a grail.
- Rarity premium is non-linear: rare usually 2-5× floor, epic 10-50×, legendary — hundreds of times.
- Different marketplaces can price the same rarity differently — that’s the rarity-arbitrage window.
How the rarity scale works
Tier systems across collections look similar but are not identical. The base scheme most upgraded-gift collections follow:
| Tier | Typical drop frequency | Description |
|---|---|---|
| Common | 4-15% | Base attribute, frequent |
| Uncommon | 1-4% | Slightly visually distinct |
| Rare | 0.5-2% | Noticeably rare, stands out in the collection |
| Epic | 0.1-0.5% | Very rare, clear collectible value |
| Legendary | <0.1% | Grail-level attribute, single-digit instances |
In a collection with 20-25 backdrop variants, typically 60-70% of total probability is spent on the common tier (5-10 backdrops), 20-25% on uncommon (5-7), 5-10% on rare (3-5), and the rest on epic+legendary (2-3 attributes). Same for symbol. Model is usually coarser: 4-6 common, 2-3 rare, 1-2 epic/legendary.
iSource of probabilities
Telegram does not officially publish probability tables. They are derived empirically — by analysing on-chain stats of minted lots. On marketplaces like MRKT this stat is embedded into the card and updates automatically.
How to read a marketplace lot card
Take a hypothetical upgraded Plush Pepe lot on MRKT. The card typically shows:
Plush Pepe #4523
Backdrop: Aquamarine — 3.2% (Rare)
Symbol: Lightning — 0.7% (Epic)
Model: Holographic — 0.4% (Epic)
Floor: 50 TON • Listed: 247 TON • Suggested: 280-350 TON
What matters here:
1. Each attribute has its own tier. Backdrop rare, symbol epic, model epic. That’s a mixed-epic combination.
2. Full lot rarity = 0.032 × 0.007 × 0.004 = ~0.0000009 = 1 in ~1.1M upgrades theoretically. In practice the collection has had N actual mints — say, 10,000 upgrades, then such lots should be fewer than 1 (likely zero).
3. Suggested price — the marketplace’s algorithmic estimate based on history of comparable sales. Not “fair value,” but a statistical anchor. Real price depends on whether a buyer for this specific lot exists.
4. Listed price — what the seller chose. Can be above or below suggested.
Rarity premium: empirical observations
Looking at historical sales in large collections, the rough pattern is (order of magnitude, not exact numbers):
| Composition tier | Floor multiplier |
|---|---|
| All three common | 1.0× (= floor) |
| One uncommon, two common | 1.2-1.5× |
| One rare, two common | 1.5-3× |
| Two rare, one common | 3-7× |
| One epic + mixed | 5-15× |
| Two epic + mixed | 20-50× |
| All three epic | 50-200× |
| Any legendary | 100-1000× |
Why non-linear: rare attributes address different buyer segments. A common lot is bought by any trader for arbitrage. A rare lot — by a collector interested in the collection. An epic+ lot — by a collector with big budget and a wish to own a “top” of their collection. Each segment pays differently.
!Premia are observed, not guaranteed
The multipliers above are an example of how the market historically priced rarity. They do not guarantee future prices. A collection can lose interest, and a grail can sell for less than what someone paid for it. Treat as observed pattern, not as price guarantee.
Type of rarity: attribute vs “figure”
A conceptual distinction frequently confused.
Attribute rarity — what we covered above: how rare a specific backdrop/symbol/model is. Applies to every upgraded gift.
“Figure” rarity — market perception of a specific series or collection as a whole. Plush Pepe, for example, has no internal attribute that makes it “legendary” in the Telegram sense — but its market cap makes it a “legendary” collection. Different rarity type.
When buying a lot you pay for both rarities simultaneously:
- Collection floor = “figure” premium of the series.
- Multiplier above floor = “attribute” premium of the specific lot.
How marketplaces display rarity
| Marketplace | Rarity display | Strengths |
|---|---|---|
| MRKT | Detailed — frequency + tier per attribute + suggested price | Best for collectors |
| Tonnel | Standard — attribute frequencies | Good for arbitrage |
| Portals | Basic — attribute icons | Most mass-market |
| Getgems | NFT-metadata-driven | NFT-centric view |
On Portals the mass trader rarely differentiates between lots finely — everything sorts by price. On MRKT you can filter by “model = Holographic AND backdrop frequency < 1%” — that is your window for rarity arbitrage.
Common mistakes when reading rarity
1. Treating attribute frequency = combination probability. If backdrop is 1% and symbol is 1%, a lot with both is 0.01%, not 1%. Multiplication is required.
2. Ignoring model. Model is usually the least likely attribute to have legendary values, but visually it is the most prominent — the market values rare models heavily.
3. Comparing rarity across different collections. Backdrop “Sapphire” in Plush Pepe might be 0.5%; “Sapphire” in another collection — 3%. Scales are independent.
4. Buying a lot “on the collection’s hype” without checking attributes. A common lot in a hot collection at hype peak is often overpriced. When hype fades it returns to floor; a rare lot holds better.
5. Ignoring marketplace suggested price. Not necessarily “correct,” but a useful sanity check. If listed >> suggested, you need to understand why.
Grails: what they are and how to handle them
A “grail” (Holy Grail) is a lot with three simultaneously epic/legendary attributes. In a large collection 1 in hundreds of thousands of upgrades; in an incomplete one — may not exist at all.
Handling a grail you own:
- Do not rush to sell. Grails are hard to liquidate quickly — thin-lot markets are thin.
- Do not list at “fair value.” List higher — big collectors want grails and pay a premium.
- Study sale history of comparable lots. Tonnel and MRKT have the deepest stats.
- Consider DAOlama. If you want liquidity but not to sell — collateralise on DAOlama for a TON loan.
Handling a grail you want to buy:
- Realistically estimate how long you can hold. Grails are multi-month positions.
- Check no comparable lots are cheaper on other marketplaces.
- Account for taxes (see tax guide).
Where attributes live on-chain
If you want to verify a lot directly via the blockchain:
- Open Tonscan or Tonviewer.
- Find the NFT token for the lot (by its address from the marketplace card).
- Metadata contains attributes — those are your three.
- Image/animation field — link to media (usually IPFS or Telegram CDN).
Useful when you want to verify a lot really matches what the marketplace shows — image-substitution scams have happened.
Practical checklist
- Studied the rarity scale of the specific collection you care about.
- Understand overall rarity = product of three attribute probabilities.
- Compare rarity across 2-3 marketplaces before buying.
- Do not buy common lots in collections at hype peak.
- Use MRKT advanced filters to hunt for underpriced rarity.
- Verify on-chain metadata before big trades.
- Remember: rarity premia are observed, not guaranteed.
Further reading
- Upgraded vs Regular Telegram gifts — base attribute mechanics.
- Gift marketplaces: where to see rarity.
- Flipping strategies — rarity arbitrage as a strategy.
- Best collections for investment — where grails hold premium.
- DAOlama: NFT-collateral lending — what to do with a grail without selling.














