Key of the silver tongue


 * 1) Customer. Avoid a 0% on every offer - you can quickly check in Item Acceptance by Class . (Game does this automatically) Except for two-handed weapons to certain classes. Musketeers do not accept fine great swords, for example. (Musketeers now take two handed swords. The same holds true for other classes that previously would not have taken two handed weapons, i.e. healers will accept deathbringers)
 * 2) Price and pocket similarity. Generally, avoid deviating from the initial item price (except with regard to token items as described below). This can result in extremely low chances of success if you deviate over 200% of the original item price. Similarly, suggesting a much lower-priced item can result in very poor chances.
 * 3) Token Item: If an item desired is one which requires tokens in order to research, then there is almost always a suggestion that is 90% likely to succeed regardless of the pocket similarity.  For example, Cow Leather Armor, with a market value of 1235 is 90% replaceable by Perfect Leather which has a value of 1910.  It appears the developers wished to design the game to not necessarily punish those who choose not to use tokens.  It also seems this is only true when the item desired is not in your inventory, i.e.; if you offer a Bone Shield to someone looking for a Hardwood Buckler when you actually have the token item, it only gives you a 55% suggestion rate; otherwise, it is 90%.  It is possible in this way to net a larger return on the asking price by suggesting the higher value item.
 * {|class="wikitable"


 * An easily estimated percentage reduction is &plusmn; $$0.5 op = 0.5 sr$$. (op original item price, sr reduction of the remainder of your success chance, by percentage) The common result of suggesting an item from a different type with half the price: $$0.5x0.5= 0.25$$
 * }
 * 1) Item category. The clearest definer of items, usually you just have to consider weapon, apparel, or consumable.
 * 2) Item affinity. Some customers will allow greater success rate for suggesting a specific item type.
 * 3) Item subcategory. A consumable can be a scroll/book, remedy, health potion, or magic potion, an accessory could be a ring, necklace, or talisman, but we know by their name and appearance which other items may be comparable. ( common mistake: Offering a necklace in place of a ring has low success, although they appear to both be jewelry. The customer wanted a subclass of ring, which would improve your chance to sell other rings. This also decreased your chance to sell a weapon, or a consumable, due to their hidden demands. )
 * 4) Weapon handedness. A one-handed and a two-handed weapon are not equal. Weapons that are two handed will usually say it under the name if you mouse over the image found with customer dialogue, or recipe ingredient. (Not true anymore, but the wiki contains data on which weapons are two-handed.)
 * 5) Antiquate. The customer who comes in and offers a sorely low bid is not expecting to win and actually makes comments which differ from normal when you say no, such as a blurry expletive, or how smart you seem. It may be easier to suggest something they like which is fast produced, and cut your loss since haggling is based on how much they like the item and how far from the mark they began, frequently a vast unknown.
 * 6) This item was close enough. (*!see below first!.) A subset within subclasses overlap for a similarity boost increasing the percentage, but just hardly enough to make the offer possible (Example: A fighter class appreciates a healing potion more when you could not sell them body armor. The chance is comparable to if they had asked for a similar potion.)
 * 7) Silver-tongue. A suggestion's success rate appears to vary as a function of the square root of silver-tongue level. Any improvements must also be considered. Note: the r-square of this function given a constant k for all pairs of requested/suggested fits with an r-square better than 0.99 in controlled tests (constant improvements, items, etc, only variant was silver-tongue points) with a large data set (though biased by a high brightest rod suggestion).


 * {|class="wikitable"

k * (requested / suggested) * sqrt(silvertongue)
 * Close enough is more of an observation flaw caused by repeated success selling an item in place of similarly priced goods. It remains here for future acknowledgment, you can sell something completely different with 40% success repeatedly without knowing anything above.
 * }
 * As a result of these points, it is reasonable to consider the following function a good estimate of success rate when suggesting an item that is more expensive than requested:
 * As a result of these points, it is reasonable to consider the following function a good estimate of success rate when suggesting an item that is more expensive than requested:
 * where k is a coefficient based on the customer's acceptance level with respect to the item suggested.
 * requested is the market-value of the item requested by the customer.
 * suggested is the market-value of the item suggested by the player.
 * silvertongue is the amount of points the player has in silvertongue, but is also modified by a player's improvements.
 * It is likely that the (requested/suggested) portion is inverted when the value of the item suggested is lower than the one requested but was not included in tests. If this is true, add an exponent to this portion.

Originally posted: ActiveUnique 05:56, July 20, 2011 (UTC)

Modified by pfb / Added statistical analysis report Jan 27, 2013